新增: DeepSeek 全部 33 篇官方论文(原文 PDF 走 Git LFS + 中文精读笔记与索引)
This commit is contained in:
ylhan
@@ -0,0 +1,60 @@
|
||||
"""Append newly discovered papers to the plan without disturbing existing numbering.
|
||||
|
||||
Usage: python add_papers.py <plan.json> <verified_extra.json> <base_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import shutil
|
||||
import sys
|
||||
|
||||
# explicit stems so existing numbered files are never renumbered
|
||||
ASSIGN = {
|
||||
'2310.16818': ('05-多模态', '08-DreamCraft3D-2310.16818', '多模态'),
|
||||
'2408.15664': ('07-架构', '04-Aux-Loss-Free-Load-Balancing-2408.15664', '架构'),
|
||||
'2411.07975': ('05-多模态', '07-JanusFlow-2411.07975', '多模态'),
|
||||
'2502.07316': ('02-代码模型', '03-CodeIO-2502.07316', '代码模型'),
|
||||
'2502.11089': ('07-架构', '03-NSA-原生稀疏注意力-2502.11089', '架构'),
|
||||
}
|
||||
|
||||
|
||||
def main():
|
||||
plan_path, ver_path, base = sys.argv[1], sys.argv[2], sys.argv[3]
|
||||
plan = json.load(open(plan_path, encoding='utf-8'))
|
||||
have = {p['id'] for p in plan}
|
||||
extra = json.load(open(ver_path, encoding='utf-8'))
|
||||
added = []
|
||||
for r in extra:
|
||||
if not r.get('ok') or r['id'] in have:
|
||||
continue
|
||||
d, stem, cat = ASSIGN[r['id']]
|
||||
outdir = os.path.join(base, d)
|
||||
os.makedirs(outdir, exist_ok=True)
|
||||
src = os.path.join(base, 'pdf', r['id'] + '.pdf')
|
||||
dst = os.path.join(outdir, stem + '.pdf')
|
||||
if os.path.exists(src) and not os.path.exists(dst):
|
||||
shutil.move(src, dst)
|
||||
plan.append({
|
||||
'id': r['id'],
|
||||
'title': r.get('title', ''),
|
||||
'authors': r.get('authors', ''),
|
||||
'abstract': r.get('abstract', ''),
|
||||
'dateline': r.get('dateline', ''),
|
||||
'date': r.get('dateline', ''),
|
||||
'comments': r.get('comments', ''),
|
||||
'category': cat,
|
||||
'dir': d,
|
||||
'stem': stem,
|
||||
'pdf_path': os.path.join(d, stem + '.pdf'),
|
||||
'md_path': os.path.join(d, stem + '.md'),
|
||||
'pdf_url': 'https://arxiv.org/pdf/' + r['id'],
|
||||
'abs_url': 'https://arxiv.org/abs/' + r['id'],
|
||||
})
|
||||
added.append((r['id'], d, stem, r.get('title', '')))
|
||||
json.dump(plan, open(plan_path, 'w', encoding='utf-8'), ensure_ascii=False, indent=1)
|
||||
for a in added:
|
||||
print('added', a)
|
||||
print('total', len(plan))
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,55 @@
|
||||
import urllib.request, urllib.parse, json, sys, time, re
|
||||
import xml.etree.ElementTree as ET
|
||||
|
||||
NS = {'a':'http://www.w3.org/2005/Atom'}
|
||||
|
||||
def fetch(q, start=0, max_results=100, sort='submittedDate'):
|
||||
params = {
|
||||
'search_query': q,
|
||||
'start': start,
|
||||
'max_results': max_results,
|
||||
'sortBy': sort,
|
||||
'sortOrder': 'descending',
|
||||
}
|
||||
url = 'http://export.arxiv.org/api/query?' + urllib.parse.urlencode(params)
|
||||
req = urllib.request.Request(url, headers={
|
||||
'User-Agent': 'DeepSeekPaperCollector/1.0 (local research notes)',
|
||||
'Accept': 'application/atom+xml',
|
||||
})
|
||||
for attempt in range(4):
|
||||
try:
|
||||
with urllib.request.urlopen(req, timeout=60) as r:
|
||||
return r.read()
|
||||
except Exception as e:
|
||||
print('retry', attempt, e, file=sys.stderr)
|
||||
time.sleep(3)
|
||||
raise SystemExit('failed')
|
||||
|
||||
def parse(xml):
|
||||
root = ET.fromstring(xml)
|
||||
out = []
|
||||
for e in root.findall('a:entry', NS):
|
||||
aid = e.find('a:id', NS).text
|
||||
m = re.search(r'arxiv.org/abs/(.+)$', aid)
|
||||
out.append({
|
||||
'id': m.group(1) if m else aid,
|
||||
'title': ' '.join(e.find('a:title', NS).text.split()),
|
||||
'published': e.find('a:published', NS).text,
|
||||
'updated': e.find('a:updated', NS).text,
|
||||
'authors': [a.find('a:name', NS).text for a in e.findall('a:author', NS)],
|
||||
'summary': ' '.join(e.find('a:summary', NS).text.split()),
|
||||
'comment': (e.find('{http://arxiv.org/schemas/atom}comment').text if e.find('{http://arxiv.org/schemas/atom}comment') is not None else ''),
|
||||
'primary': (e.find('{http://arxiv.org/schemas/atom}primary_category').get('term') if e.find('{http://arxiv.org/schemas/atom}primary_category') is not None else ''),
|
||||
})
|
||||
return out
|
||||
|
||||
if __name__ == '__main__':
|
||||
query = sys.argv[1]
|
||||
allr = []
|
||||
for start in range(0, 400, 100):
|
||||
xml = fetch(query, start=start)
|
||||
res = parse(xml)
|
||||
allr.extend(res)
|
||||
if len(res) < 100:
|
||||
break
|
||||
print(json.dumps(allr, ensure_ascii=False, indent=1))
|
||||
@@ -0,0 +1,50 @@
|
||||
[
|
||||
{
|
||||
"id": "2609.19969",
|
||||
"title": "DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression",
|
||||
"authors": "Authors: DeepSeek-AI , : , Anyi Xu , B. Li , Bangcai Lin , Bing Xue , BingCheng Xian , Bingzheng Xu , Bochao Wu , Bowei Zhang , Boyi Deng , C. C. Yu , Chao Jin , Chaofan Lin , Chen Dong , Chenbing Wang , Chenfan Feng , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chengyuan Zhang , Chenhao Xu , Chenqi Zhao , Chenze Shao , Chuhao Wang , et al. (568 additional authors not shown)",
|
||||
"dates": "Submitted 17 September, 2026; originally announced September 2026."
|
||||
},
|
||||
{
|
||||
"id": "2606.19348",
|
||||
"title": "DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence",
|
||||
"authors": "Authors: DeepSeek-AI , Anyi Xu , Bangcai Lin , Bing Xue , Bingxuan Wang , Bingzheng Xu , Bochao Wu , Bowei Zhang , Chaofan Lin , Chen Dong , Chenchen Ling , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chengyu Hou , Chenhao Xu , Chenze Shao , Chong Ruan , Conner Sun , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Donghao Li , Dongjie Ji , et al. (294 additional authors not shown)",
|
||||
"dates": "Submitted 26 April, 2026; originally announced June 2026."
|
||||
},
|
||||
{
|
||||
"id": "2512.02556",
|
||||
"title": "DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models",
|
||||
"authors": "Authors: DeepSeek-AI , Aixin Liu , Aoxue Mei , Bangcai Lin , Bing Xue , Bingxuan Wang , Bingzheng Xu , Bochao Wu , Bowei Zhang , Chaofan Lin , Chen Dong , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chenhao Xu , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Erhang Li , Fangqi Zhou , Fangyun Lin , Fucong Dai , Guangbo Hao , et al. (239 additional authors not shown)",
|
||||
"dates": "Submitted 2 December, 2025; originally announced December 2025."
|
||||
},
|
||||
{
|
||||
"id": "2501.12948",
|
||||
"title": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning",
|
||||
"authors": "Authors: DeepSeek-AI , Daya Guo , Dejian Yang , Haowei Zhang , Junxiao Song , Peiyi Wang , Qihao Zhu , Runxin Xu , Ruoyu Zhang , Shirong Ma , Xiao Bi , Xiaokang Zhang , Xingkai Yu , Yu Wu , Z. F. Wu , Zhibin Gou , Zhihong Shao , Zhuoshu Li , Ziyi Gao , Aixin Liu , Bing Xue , Bingxuan Wang , Bochao Wu , Bei Feng , Chengda Lu , et al. (175 additional authors not shown)",
|
||||
"dates": "Submitted 3 January, 2026; v1 submitted 22 January, 2025; originally announced January 2025."
|
||||
},
|
||||
{
|
||||
"id": "2412.19437",
|
||||
"title": "DeepSeek-V3 Technical Report",
|
||||
"authors": "Authors: DeepSeek-AI , Aixin Liu , Bei Feng , Bing Xue , Bingxuan Wang , Bochao Wu , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chenyu Zhang , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Dongjie Ji , Erhang Li , Fangyun Lin , Fucong Dai , Fuli Luo , Guangbo Hao , Guanting Chen , Guowei Li , H. Zhang , Han Bao , et al. (175 additional authors not shown)",
|
||||
"dates": "Submitted 18 February, 2025; v1 submitted 26 December, 2024; originally announced December 2024."
|
||||
},
|
||||
{
|
||||
"id": "2406.11931",
|
||||
"title": "DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence",
|
||||
"authors": "Authors: DeepSeek-AI , Qihao Zhu , Daya Guo , Zhihong Shao , Dejian Yang , Peiyi Wang , Runxin Xu , Y. Wu , Yukun Li , Huazuo Gao , Shirong Ma , Wangding Zeng , Xiao Bi , Zihui Gu , Hanwei Xu , Damai Dai , Kai Dong , Liyue Zhang , Yishi Piao , Zhibin Gou , Zhenda Xie , Zhewen Hao , Bingxuan Wang , Junxiao Song , Deli Chen , et al. (15 additional authors not shown)",
|
||||
"dates": "Submitted 17 June, 2024; originally announced June 2024."
|
||||
},
|
||||
{
|
||||
"id": "2405.04434",
|
||||
"title": "DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model",
|
||||
"authors": "Authors: DeepSeek-AI , Aixin Liu , Bei Feng , Bin Wang , Bingxuan Wang , Bo Liu , Chenggang Zhao , Chengqi Dengr , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Dongjie Ji , Erhang Li , Fangyun Lin , Fuli Luo , Guangbo Hao , Guanting Chen , Guowei Li , H. Zhang , Hanwei Xu , Hao Yang , Haowei Zhang , Honghui Ding , et al. (132 additional authors not shown)",
|
||||
"dates": "Submitted 19 June, 2024; v1 submitted 7 May, 2024; originally announced May 2024."
|
||||
},
|
||||
{
|
||||
"id": "2401.02954",
|
||||
"title": "DeepSeek LLM: Scaling Open-Source Language Models with Longtermism",
|
||||
"authors": "Authors: DeepSeek-AI , : , Xiao Bi , Deli Chen , Guanting Chen , Shanhuang Chen , Damai Dai , Chengqi Deng , Honghui Ding , Kai Dong , Qiushi Du , Zhe Fu , Huazuo Gao , Kaige Gao , Wenjun Gao , Ruiqi Ge , Kang Guan , Daya Guo , Jianzhong Guo , Guangbo Hao , Zhewen Hao , Ying He , Wenjie Hu , Panpan Huang , Erhang Li , et al. (63 additional authors not shown)",
|
||||
"dates": "Submitted 5 January, 2024; originally announced January 2024."
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,8 @@
|
||||
2406.11931 DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Cod Submitted 17 June, 2024; originally an
|
||||
2609.19969 DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression Submitted 17 September, 2026; original
|
||||
2412.19437 DeepSeek-V3 Technical Report Submitted 18 February, 2025; v1 submit
|
||||
2405.04434 DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts La Submitted 19 June, 2024; v1 submitted
|
||||
2512.02556 DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models Submitted 2 December, 2025; originally
|
||||
2606.19348 DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligen Submitted 26 April, 2026; originally a
|
||||
2501.12948 DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcem Submitted 3 January, 2026; v1 submitte
|
||||
2401.02954 DeepSeek LLM: Scaling Open-Source Language Models with Longtermism Submitted 5 January, 2024; originally
|
||||
@@ -0,0 +1,90 @@
|
||||
"""Merge verified arXiv metadata into a master paper list.
|
||||
|
||||
Usage: python build_list.py <out.json> <verified1.json> <verified2.json> ...
|
||||
"""
|
||||
import json
|
||||
import sys
|
||||
import re
|
||||
import os
|
||||
|
||||
OUT = sys.argv[1]
|
||||
# verified IDs that turned out NOT to be DeepSeek papers
|
||||
SKIP = {'2411.08599', '2605.27765'}
|
||||
recs = {}
|
||||
for p in sys.argv[2:]:
|
||||
if not os.path.exists(p):
|
||||
continue
|
||||
with open(p, encoding='utf-8') as f:
|
||||
for r in json.load(f):
|
||||
if r.get('ok') and r['id'] not in SKIP:
|
||||
recs[r['id']] = r
|
||||
|
||||
# group/category assignment
|
||||
CATS = {
|
||||
'2401.02954': '基础模型',
|
||||
'2401.06066': '基础模型',
|
||||
'2401.14196': '代码模型',
|
||||
'2402.03300': '数学模型',
|
||||
'2403.05525': '多模态',
|
||||
'2405.04434': '基础模型',
|
||||
'2405.14333': '形式化证明',
|
||||
'2406.11931': '代码模型',
|
||||
'2407.01906': '训练方法',
|
||||
'2408.08152': '形式化证明',
|
||||
'2408.14158': '基础设施',
|
||||
'2410.13848': '多模态',
|
||||
'2412.10302': '多模态',
|
||||
'2412.19437': '基础模型',
|
||||
'2501.12948': '推理模型',
|
||||
'2501.17811': '多模态',
|
||||
'2504.02495': '训练方法',
|
||||
'2504.21801': '形式化证明',
|
||||
'2505.09343': '基础设施',
|
||||
'2510.18234': '多模态',
|
||||
'2511.22570': '数学模型',
|
||||
'2512.02556': '基础模型',
|
||||
'2512.24880': '架构',
|
||||
'2601.07372': '架构',
|
||||
'2601.20552': '多模态',
|
||||
'2602.21548': '基础设施',
|
||||
'2606.19348': '基础模型',
|
||||
'2609.19969': '基础模型',
|
||||
}
|
||||
|
||||
|
||||
def slug(title, aid):
|
||||
s = title.lower()
|
||||
s = re.sub(r'[^a-z0-9]+', '-', s).strip('-')
|
||||
s = re.sub(r'-+', '-', s)
|
||||
return (s[:70] or aid) + '-' + aid.replace('.', '')
|
||||
|
||||
|
||||
rows = []
|
||||
for aid, r in recs.items():
|
||||
m = re.search(r'(\d{4})', r.get('dateline', ''))
|
||||
date = ''
|
||||
dm = re.search(r'(\d{1,2} \w+ \d{4})', r.get('dateline', ''))
|
||||
if dm:
|
||||
date = dm.group(1)
|
||||
rows.append({
|
||||
'id': aid,
|
||||
'slug': slug(r.get('title', ''), aid),
|
||||
'title': r.get('title', ''),
|
||||
'authors': r.get('authors', ''),
|
||||
'abstract': r.get('abstract', ''),
|
||||
'date': date,
|
||||
'dateline': r.get('dateline', ''),
|
||||
'comments': r.get('comments', ''),
|
||||
'category': CATS.get(aid, '其他'),
|
||||
'pdf_url': 'https://arxiv.org/pdf/' + aid,
|
||||
'abs_url': 'https://arxiv.org/abs/' + aid,
|
||||
})
|
||||
rows.sort(key=lambda x: x['id'])
|
||||
groups = {}
|
||||
for r in rows:
|
||||
groups.setdefault(r['category'], []).append(r)
|
||||
with open(OUT, 'w', encoding='utf-8') as f:
|
||||
json.dump(rows, f, ensure_ascii=False, indent=1)
|
||||
print('papers:', len(rows))
|
||||
for c, v in sorted(groups.items()):
|
||||
print(' ', c, len(v))
|
||||
@@ -0,0 +1,27 @@
|
||||
# DeepSeek official paper candidate IDs
|
||||
2401.02954 # DeepSeek LLM
|
||||
2401.06066 # DeepSeekMoE
|
||||
2401.14196 # DeepSeek-Coder
|
||||
2402.03300 # DeepSeekMath
|
||||
2403.05525 # DeepSeek-VL
|
||||
2405.04434 # DeepSeek-V2
|
||||
2405.14333 # DeepSeek-Prover
|
||||
2406.11931 # DeepSeek-Coder-V2
|
||||
2408.08152 # DeepSeek-Prover-V1.5
|
||||
2408.14158 # Fire-Flyer AI-HPC
|
||||
2410.13848 # Janus
|
||||
2411.08599 # ESFT
|
||||
2412.10302 # DeepSeek-VL2
|
||||
2412.19437 # DeepSeek-V3
|
||||
2501.12948 # DeepSeek-R1
|
||||
2501.17811 # Janus-Pro
|
||||
2504.21801 # DeepSeek-Prover-V2
|
||||
2505.09343 # Insights into DeepSeek-V3
|
||||
2510.18234 # DeepSeek-OCR
|
||||
2511.22570 # DeepSeekMath-V2
|
||||
2512.02556 # DeepSeek-V3.2
|
||||
2601.07372 # Conditional Memory via Scalable Lookup (Engram)
|
||||
2601.20552 # DeepSeek-OCR 2
|
||||
2605.27765 # unknown - check
|
||||
2606.19348 # DeepSeek-V4
|
||||
2609.19969 # DeepSeek-V4.1-Flash
|
||||
@@ -0,0 +1,2 @@
|
||||
2407.01906 # ESFT: Let the Expert Stick to His Last
|
||||
2504.02495 # Inference-Time Scaling for Generalist Reward Modeling
|
||||
@@ -0,0 +1 @@
|
||||
2512.24880
|
||||
@@ -0,0 +1 @@
|
||||
2602.21548
|
||||
@@ -0,0 +1,5 @@
|
||||
2310.16818
|
||||
2408.15664
|
||||
2411.07975
|
||||
2502.07316
|
||||
2502.11089
|
||||
@@ -0,0 +1,105 @@
|
||||
"""Collect DeepSeek papers from the arXiv API.
|
||||
|
||||
Usage: python collect.py <output.json>
|
||||
"""
|
||||
import urllib.request
|
||||
import urllib.parse
|
||||
import json
|
||||
import sys
|
||||
import time
|
||||
import re
|
||||
import os
|
||||
import xml.etree.ElementTree as ET
|
||||
|
||||
NS = {'a': 'http://www.w3.org/2005/Atom',
|
||||
'arxiv': 'http://arxiv.org/schemas/atom'}
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def fetch(q, start=0, max_results=100, sort='submittedDate'):
|
||||
params = {
|
||||
'search_query': q,
|
||||
'start': start,
|
||||
'max_results': max_results,
|
||||
'sortBy': sort,
|
||||
'sortOrder': 'descending',
|
||||
}
|
||||
url = 'https://export.arxiv.org/api/query?' + urllib.parse.urlencode(params)
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
last = None
|
||||
for attempt in range(5):
|
||||
try:
|
||||
with urllib.request.urlopen(req, timeout=90) as r:
|
||||
return r.read()
|
||||
except Exception as e: # noqa: BLE001
|
||||
last = e
|
||||
time.sleep(4)
|
||||
print('FETCH FAILED', q, last, file=sys.stderr)
|
||||
return None
|
||||
|
||||
|
||||
def parse(xml):
|
||||
root = ET.fromstring(xml)
|
||||
out = []
|
||||
for e in root.findall('a:entry', NS):
|
||||
aid = e.find('a:id', NS).text
|
||||
m = re.search(r'arxiv\.org/abs/(.+)$', aid)
|
||||
comment = e.find('arxiv:comment', NS)
|
||||
primary = e.find('arxiv:primary_category', NS)
|
||||
journal = e.find('arxiv:journal_ref', NS)
|
||||
doi = e.find('arxiv:doi', NS)
|
||||
out.append({
|
||||
'id': (m.group(1) if m else aid),
|
||||
'version_id': aid,
|
||||
'title': ' '.join(e.find('a:title', NS).text.split()),
|
||||
'published': e.find('a:published', NS).text,
|
||||
'updated': e.find('a:updated', NS).text,
|
||||
'authors': [a.find('a:name', NS).text for a in e.findall('a:author', NS)],
|
||||
'summary': ' '.join(e.find('a:summary', NS).text.split()),
|
||||
'comment': (' '.join(comment.text.split()) if comment is not None and comment.text else ''),
|
||||
'journal_ref': (' '.join(journal.text.split()) if journal is not None and journal.text else ''),
|
||||
'doi': (doi.text.strip() if doi is not None and doi.text else ''),
|
||||
'primary': (primary.get('term') if primary is not None else ''),
|
||||
})
|
||||
return out
|
||||
|
||||
|
||||
def search(q, cap=200):
|
||||
allr = []
|
||||
for start in range(0, cap, 100):
|
||||
xml = fetch(q, start=start)
|
||||
if not xml:
|
||||
break
|
||||
res = parse(xml)
|
||||
allr.extend(res)
|
||||
if len(res) < 100:
|
||||
break
|
||||
time.sleep(3)
|
||||
return allr
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
queries = [
|
||||
'all:"DeepSeek"',
|
||||
'abs:"DeepSeek"',
|
||||
'ti:"DeepSeek"',
|
||||
'all:"DeepSeek-V3" OR all:"DeepSeek-R1"',
|
||||
'all:"DeepSeekMath" OR all:"DeepSeek-Coder" OR all:"DeepSeek-VL"',
|
||||
'all:"DeepSeekMoE" OR all:"DeepSeek-Prover" OR all:"Janus"',
|
||||
'all:"Fire-Flyer" OR all:"DeepSeek-OCR"',
|
||||
]
|
||||
seen = {}
|
||||
for q in queries:
|
||||
print('query:', q, file=sys.stderr)
|
||||
for r in search(q):
|
||||
key = r['id'].split('v')[0]
|
||||
if key not in seen:
|
||||
seen[key] = r
|
||||
time.sleep(3)
|
||||
res = sorted(seen.values(), key=lambda x: x['published'])
|
||||
out = sys.argv[1] if len(sys.argv) > 1 else 'raw_deepseek.json'
|
||||
os.makedirs(os.path.dirname(out), exist_ok=True)
|
||||
with open(out, 'w', encoding='utf-8') as f:
|
||||
json.dump(res, f, ensure_ascii=False, indent=1)
|
||||
print('total', len(res), file=sys.stderr)
|
||||
@@ -0,0 +1,27 @@
|
||||
"""Print HTML context around a keyword.
|
||||
|
||||
Usage: python ctx.py <url> <keyword> [window]
|
||||
"""
|
||||
import urllib.request
|
||||
import sys
|
||||
import re
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
url = sys.argv[1]
|
||||
kw = sys.argv[2]
|
||||
win = int(sys.argv[3]) if len(sys.argv) > 3 else 600
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
t = urllib.request.urlopen(req, timeout=40).read().decode('utf-8', 'ignore')
|
||||
idx = 0
|
||||
n = 0
|
||||
while True:
|
||||
i = t.find(kw, idx)
|
||||
if i < 0:
|
||||
break
|
||||
n += 1
|
||||
print('--- occurrence %d at %d ---' % (n, i))
|
||||
print(t[max(0, i - win):i + win])
|
||||
idx = i + 1
|
||||
if n >= 5:
|
||||
break
|
||||
print('occurrences found:', n)
|
||||
@@ -0,0 +1,57 @@
|
||||
"""Download arXiv PDFs for the master paper list.
|
||||
|
||||
Usage: python download.py <papers.json> <pdf_dir>
|
||||
Idempotent: skips files that already exist with a plausible size.
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import sys
|
||||
import time
|
||||
import urllib.request
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
MIN_BYTES = 20000
|
||||
|
||||
|
||||
def fetch(url, timeout=180):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read()
|
||||
|
||||
|
||||
def main():
|
||||
papers = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
outdir = sys.argv[2]
|
||||
os.makedirs(outdir, exist_ok=True)
|
||||
ok = 0
|
||||
fail = []
|
||||
for p in papers:
|
||||
dest = os.path.join(outdir, p['id'] + '.pdf')
|
||||
if os.path.exists(dest) and os.path.getsize(dest) > MIN_BYTES:
|
||||
print('SKIP', p['id'], os.path.getsize(dest))
|
||||
ok += 1
|
||||
continue
|
||||
got = False
|
||||
for attempt in range(3):
|
||||
try:
|
||||
data = fetch(p['pdf_url'])
|
||||
if data[:4] == b'%PDF' and len(data) > MIN_BYTES:
|
||||
with open(dest, 'wb') as f:
|
||||
f.write(data)
|
||||
print('OK ', p['id'], len(data), p['title'][:60])
|
||||
ok += 1
|
||||
got = True
|
||||
break
|
||||
print(' bad payload', p['id'], len(data), data[:20])
|
||||
except Exception as e: # noqa: BLE001
|
||||
print(' retry', attempt, p['id'], str(e)[:100])
|
||||
time.sleep(3)
|
||||
if not got:
|
||||
fail.append(p['id'])
|
||||
print('FAIL', p['id'])
|
||||
time.sleep(1)
|
||||
print('=== downloaded/verified %d/%d, failures: %s' % (ok, len(papers), fail))
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,45 @@
|
||||
"""Download PDFs by arXiv id list.
|
||||
|
||||
Usage: python download_ids.py <ids.txt> <out_dir>
|
||||
"""
|
||||
import os
|
||||
import sys
|
||||
import time
|
||||
import urllib.request
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
MIN = 20000
|
||||
|
||||
|
||||
def main():
|
||||
ids = [l.strip() for l in open(sys.argv[1], encoding='utf-8') if l.strip()]
|
||||
outdir = sys.argv[2]
|
||||
os.makedirs(outdir, exist_ok=True)
|
||||
fail = []
|
||||
for aid in ids:
|
||||
dest = os.path.join(outdir, aid + '.pdf')
|
||||
if os.path.exists(dest) and os.path.getsize(dest) > MIN:
|
||||
print('SKIP', aid, os.path.getsize(dest))
|
||||
continue
|
||||
done = False
|
||||
for attempt in range(4):
|
||||
try:
|
||||
req = urllib.request.Request('https://arxiv.org/pdf/' + aid,
|
||||
headers={'User-Agent': UA})
|
||||
data = urllib.request.urlopen(req, timeout=180).read()
|
||||
if data[:4] == b'%PDF' and len(data) > MIN:
|
||||
open(dest, 'wb').write(data)
|
||||
print('OK ', aid, len(data))
|
||||
done = True
|
||||
break
|
||||
print(' bad', aid, len(data))
|
||||
except Exception as e: # noqa: BLE001
|
||||
print(' retry', attempt, aid, str(e)[:90])
|
||||
time.sleep(3)
|
||||
if not done:
|
||||
fail.append(aid)
|
||||
print('failures:', fail)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,31 @@
|
||||
"""Dump all links found on a page.
|
||||
|
||||
Usage: python dump_links.py <url> [filter]
|
||||
"""
|
||||
import urllib.request
|
||||
import re
|
||||
import sys
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def get(url, timeout=40):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
url = sys.argv[1]
|
||||
filt = sys.argv[2] if len(sys.argv) > 2 else ''
|
||||
t = get(url)
|
||||
links = re.findall(r'href="([^"]+)"', t)
|
||||
seen = []
|
||||
for l in links:
|
||||
if filt and filt not in l:
|
||||
continue
|
||||
if l not in seen:
|
||||
seen.append(l)
|
||||
print('TOTAL LINKS:', len(links), 'SHOWN:', len(seen))
|
||||
for l in seen:
|
||||
print(l)
|
||||
@@ -0,0 +1,13 @@
|
||||
import json
|
||||
import sys
|
||||
|
||||
src = sys.argv[1]
|
||||
dst = sys.argv[2]
|
||||
with open(src, encoding='utf-8') as f:
|
||||
rows = json.load(f)
|
||||
lines = []
|
||||
for r in sorted(rows, key=lambda x: x.get('dates', '')):
|
||||
lines.append('%-14s %-70s %s' % (r['id'], r.get('title', '')[:70], r.get('dates', '')[:38]))
|
||||
with open(dst, 'w', encoding='utf-8') as f:
|
||||
f.write('\n'.join(lines))
|
||||
print('wrote', len(lines))
|
||||
@@ -0,0 +1,105 @@
|
||||
"""Extract clean text + a digest for each paper from arXiv HTML.
|
||||
|
||||
Usage: python extract_text.py <papers.json> <text_dir> <digest_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
import time
|
||||
import html as htmlmod
|
||||
import urllib.request
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def fetch(url, timeout=90):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
def html_to_text(t):
|
||||
t = re.sub(r'<script.*?</script>', ' ', t, flags=re.S | re.I)
|
||||
t = re.sub(r'<style.*?</style>', ' ', t, flags=re.S | re.I)
|
||||
t = re.sub(r'<!--.*?-->', ' ', t, flags=re.S)
|
||||
# headings
|
||||
t = re.sub(r'<h([1-6])[^>]*>(.*?)</h\1>', lambda m: '\n\n## ' + re.sub(r'<[^>]+>', ' ', m.group(2)) + '\n', t, flags=re.S | re.I)
|
||||
# block elements -> newline
|
||||
t = re.sub(r'</(p|div|li|tr|table|section|figure|figcaption|blockquote)>', '\n', t, flags=re.I)
|
||||
t = re.sub(r'<br\s*/?>', '\n', t, flags=re.I)
|
||||
t = re.sub(r'<[^>]+>', ' ', t)
|
||||
t = htmlmod.unescape(t)
|
||||
t = t.replace('\u00a0', ' ')
|
||||
lines = []
|
||||
for ln in t.split('\n'):
|
||||
ln = ' '.join(ln.split())
|
||||
if ln:
|
||||
lines.append(ln)
|
||||
return '\n'.join(lines)
|
||||
|
||||
|
||||
def main():
|
||||
papers = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
tdir, ddir = sys.argv[2], sys.argv[3]
|
||||
os.makedirs(tdir, exist_ok=True)
|
||||
os.makedirs(ddir, exist_ok=True)
|
||||
report = []
|
||||
for p in papers:
|
||||
aid = p['id']
|
||||
tpath = os.path.join(tdir, aid + '.txt')
|
||||
dpath = os.path.join(ddir, aid + '.txt')
|
||||
if os.path.exists(tpath) and os.path.getsize(tpath) > 5000:
|
||||
report.append((aid, 'cached', os.path.getsize(tpath)))
|
||||
continue
|
||||
txt = None
|
||||
for suffix in ('', 'v1', 'v2', 'v3'):
|
||||
try:
|
||||
h = fetch('https://arxiv.org/html/%s%s' % (aid, suffix))
|
||||
if 'No HTML' in h[:5000] or len(h) < 5000:
|
||||
continue
|
||||
txt = html_to_text(h)
|
||||
if len(txt) > 5000:
|
||||
break
|
||||
txt = None
|
||||
except Exception: # noqa: BLE001
|
||||
time.sleep(2)
|
||||
if not txt:
|
||||
report.append((aid, 'NO-HTML', 0))
|
||||
continue
|
||||
with open(tpath, 'w', encoding='utf-8') as f:
|
||||
f.write(txt)
|
||||
# digest: everything before References, capped
|
||||
m = re.search(r'\n## .{0,40}References', txt, re.I)
|
||||
body = txt[:m.start()] if m else txt
|
||||
head = body[:60000]
|
||||
tail = body[-14000:] if len(body) > 74000 else body[60000:]
|
||||
headings = [l for l in body.split('\n') if l.startswith('## ')]
|
||||
digest = '\n'.join([
|
||||
'TITLE: ' + p['title'],
|
||||
'ARXIV: ' + aid,
|
||||
'AUTHORS: ' + p['authors'][:1500],
|
||||
'DATE: ' + p['dateline'],
|
||||
'',
|
||||
'=== SECTION HEADINGS ===',
|
||||
'\n'.join(headings[:120]),
|
||||
'',
|
||||
'=== BODY (head) ===',
|
||||
head,
|
||||
'',
|
||||
'=== BODY (tail) ===',
|
||||
tail,
|
||||
])
|
||||
with open(dpath, 'w', encoding='utf-8') as f:
|
||||
f.write(digest)
|
||||
report.append((aid, 'ok', len(txt)))
|
||||
print('ok ', aid, len(txt), 'digest', len(digest))
|
||||
time.sleep(1)
|
||||
for r in report:
|
||||
print(r)
|
||||
bad = [r[0] for r in report if r[1] == 'NO-HTML']
|
||||
print('NO-HTML:', bad)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,19 @@
|
||||
import json
|
||||
import sys
|
||||
import re
|
||||
|
||||
p = sys.argv[1]
|
||||
with open(p, encoding='utf-8') as f:
|
||||
rows = json.load(f)
|
||||
print('rows:', len(rows))
|
||||
hits = []
|
||||
for r in rows:
|
||||
a = r.get('authors', '')
|
||||
if re.search(r'DeepSeek', a) or 'DeepSeek' in r.get('title', ''):
|
||||
hits.append(r)
|
||||
print('hits:', len(hits))
|
||||
for r in hits:
|
||||
print('-' * 90)
|
||||
print(r['id'], '|', r.get('dates', '')[:60])
|
||||
print('T:', r.get('title', ''))
|
||||
print('A:', r.get('authors', '')[:200])
|
||||
@@ -0,0 +1,81 @@
|
||||
"""Strict final check of the DeepSeek paper collection.
|
||||
|
||||
Usage: python final_check.py <plan.json> <base_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
|
||||
plan = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
base = sys.argv[2]
|
||||
problems = []
|
||||
tot_cjk = 0
|
||||
tot_pdf = 0
|
||||
|
||||
for p in plan:
|
||||
tag = p['id']
|
||||
pdf = os.path.join(base, p['pdf_path'])
|
||||
md = os.path.join(base, p['md_path'])
|
||||
# PDF
|
||||
if not os.path.exists(pdf):
|
||||
problems.append((tag, 'PDF missing'))
|
||||
else:
|
||||
with open(pdf, 'rb') as f:
|
||||
head = f.read(5)
|
||||
size = os.path.getsize(pdf)
|
||||
tot_pdf += size
|
||||
if head[:4] != b'%PDF':
|
||||
problems.append((tag, 'PDF header invalid: %r' % head))
|
||||
if size < 20000:
|
||||
problems.append((tag, 'PDF too small: %d' % size))
|
||||
# MD
|
||||
if not os.path.exists(md):
|
||||
problems.append((tag, 'MD missing'))
|
||||
continue
|
||||
t = open(md, encoding='utf-8').read()
|
||||
cjk = len(re.findall(r'[\u4e00-\u9fff]', t))
|
||||
tot_cjk += cjk
|
||||
if not t.lstrip().startswith('# '):
|
||||
problems.append((tag, 'no H1 heading'))
|
||||
for sec in ['## 一、', '## 二、', '## 三、', '## 四、', '## 五、', '## 六、']:
|
||||
if sec not in t:
|
||||
problems.append((tag, 'missing section %s' % sec))
|
||||
if cjk < 1200:
|
||||
problems.append((tag, 'too little Chinese text: %d' % cjk))
|
||||
if p['id'] not in t:
|
||||
problems.append((tag, 'arXiv id not referenced in MD'))
|
||||
if 'arxiv.org/abs/' + p['id'] not in t:
|
||||
problems.append((tag, 'abs link missing in MD'))
|
||||
# relative pdf link resolves
|
||||
m = re.search(r'\]\(\./([^)]+\.pdf)\)', t)
|
||||
if not m:
|
||||
problems.append((tag, 'no relative PDF link'))
|
||||
else:
|
||||
target = os.path.join(os.path.dirname(md), m.group(1))
|
||||
if not os.path.exists(target):
|
||||
problems.append((tag, 'relative PDF link broken: %s' % m.group(1)))
|
||||
elif target != pdf.replace('/', os.sep) and os.path.basename(target) != os.path.basename(pdf):
|
||||
problems.append((tag, 'PDF link points elsewhere: %s' % m.group(1)))
|
||||
|
||||
# README
|
||||
rd = os.path.join(base, 'README.md')
|
||||
if not os.path.exists(rd):
|
||||
problems.append(('README', 'missing'))
|
||||
else:
|
||||
rt = open(rd, encoding='utf-8').read()
|
||||
if '（待生成）' in rt:
|
||||
problems.append(('README', 'still contains 待生成 placeholder'))
|
||||
listed = len(re.findall(r'\| \d+ \| \[', rt))
|
||||
if listed != len(plan):
|
||||
problems.append(('README', 'overview rows %d != %d' % (listed, len(plan))))
|
||||
for p in plan:
|
||||
if p['stem'] not in rt:
|
||||
problems.append(('README', 'missing entry ' + p['stem']))
|
||||
|
||||
print('papers checked :', len(plan))
|
||||
print('total PDF bytes : %.1f MB' % (tot_pdf / 1024 / 1024))
|
||||
print('total Chinese chars :', tot_cjk)
|
||||
print('problems :', len(problems))
|
||||
for t, m in problems:
|
||||
print(' !', t, m)
|
||||
@@ -0,0 +1,152 @@
|
||||
DeepSeek Research Papers: V4 to V2 Reading Guide Skip to content No results Home Models Tutorials Comparisons Guides Reviews API Tools About DeepSeek AI Guide H
|
||||
ome Models Tutorials Comparisons Guides Reviews API Tools About DeepSeek AI Guide Search Menu A Guide to DeepSeek Research Papers From V2 to V4 Map every major
|
||||
DeepSeek research paper from V2 to the V4 technical report, with key contributions and benchmarks. Start reading today. April 24, 2026 News Quick facts Quick fa
|
||||
cts Topic A Guide to DeepSeek Research Papers From V2 to V4 Format News brief Models discussed V4-Pro, V4-Flash, V4, V3.2 Reading time 11 min read (~2,519 words
|
||||
) Last reviewed: April 30, 2026 · Source · View sources On this page 12 sections Why DeepSeek’s papers matter more than most The DeepSeek paper timeline at a gl
|
||||
ance The foundation papers (2024) DeepSeek LLM — the scaling-law study DeepSeekMoE and DeepSeek-V2 — the architecture base DeepSeek-Coder and DeepSeekMath DeepS
|
||||
eek V3 — the efficiency thesis in paper form DeepSeek R1 — reasoning without human demonstrations DeepSeek V3.2 — the sparse attention bridge DeepSeek V4 — the
|
||||
current paper Architecture contributions Positioning against closed frontier models How the papers connect: a suggested reading order Using the papers as a deve
|
||||
loper, not a reader What the papers do not tell you Frequently asked questions Where can I find the official DeepSeek research papers? What does the DeepSeek V4
|
||||
paper actually introduce? How much did it cost to train DeepSeek V3 and R1? Is the DeepSeek R1 paper really a reasoning trace breakthrough? Can I reproduce the
|
||||
benchmark numbers in the papers using the hosted API? Sources & Methodology Primary sources Technical references Context sources Methodology Data confidence Ed
|
||||
itorial note A Guide to DeepSeek Research Papers From V2 to V4 News · April 24, 2026 · By DS Guide Editorial If you are trying to understand why DeepSeek V4 use
|
||||
s only 10% of the KV cache of V3.2, or how a Chinese lab trained a reasoning model for under $300,000, the answer sits inside the DeepSeek research papers — not
|
||||
the press coverage. This guide maps every major publication the lab has put on arXiv and Hugging Face, from the original DeepSeek LLM scaling-law work in Janua
|
||||
ry 2024 to the DeepSeek-V4 technical report released on April 24, 2026. For each paper we note the model it describes, the single architectural idea it contribu
|
||||
ted, and how those ideas stack into the V4 release. You will finish with a reading order, a benchmark cross-reference, and pointers to our deeper model-by-model
|
||||
breakdowns. Why DeepSeek’s papers matter more than most Most frontier AI labs ship products and a blog post. DeepSeek ships products, open weights, and a detai
|
||||
led technical report — usually on arXiv, usually the same day, usually with enough implementation detail to reproduce the architecture. That is how the lab beca
|
||||
me a reference point in 2025: DeepSeek drew wide attention in 2025 when its R1 release delivered near industry-leading reasoning performance – for allegedly a f
|
||||
raction of the price. The V4 preview, released on April 24, 2026, continues that pattern. The paper presents two Mixture-of-Experts language models — DeepSeek-V
|
||||
4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) — both supporting a context length of one million tokens.
|
||||
If you want to understand what V4 actually does differently, you read the V4 technical report. This article is a reading guide, not a replacement for those prim
|
||||
ary sources. The DeepSeek paper timeline at a glance Every paper below is first-party DeepSeek-AI work. I have limited the list to the releases that matter for
|
||||
understanding the current V4 generation. For a wider bibliography, the community-maintained DeepSeek Papers collection on Hugging Face is the most complete sing
|
||||
le index. Paper Published arXiv / host Core contribution DeepSeek LLM 2024-01-05 arXiv 2401.02954 Scaling-law study for 7B and 67B dense models DeepSeekMoE 2024
|
||||
-01-11 arXiv 2401.06066 Fine-grained expert specialisation DeepSeek-Coder 2024-01-25 arXiv 2401.14196 Code-specialised base models DeepSeekMath (GRPO) 2024-02-0
|
||||
5 arXiv 2402.03300 Group Relative Policy Optimisation DeepSeek-V2 2024-05-07 arXiv 2405.04434 Multi-head Latent Attention (MLA) DeepSeek-V3 2024-12-27 arXiv 241
|
||||
2.19437 Auxiliary-loss-free load balancing, FP8 training DeepSeek-R1 2025-01-22 arXiv 2501.12948 RL-only reasoning training DeepSeek-V3.2 2025-12-02 arXiv 2512.
|
||||
02556 DeepSeek Sparse Attention (DSA) DeepSeek-V4 Preview 2026-04-24 Hugging Face PDF Hybrid CSA + HCA attention, 1M context by default The foundation papers (2
|
||||
024) DeepSeek LLM — the scaling-law study The first DeepSeek paper is not about a frontier model. It is a scaling-law study that justifies everything that came
|
||||
after. The authors argue that prior scaling literature gave conflicting guidance, so they re-derived the laws for their own training stack and applied them to 7
|
||||
B and 67B dense configurations. The published corpus at the time was 2 trillion tokens and growing. If you want context on why later DeepSeek models look the wa
|
||||
y they do — and why the lab keeps publishing token-count figures prominently — this is the origin. For a plain-English summary, see our page on DeepSeek LLM . D
|
||||
eepSeekMoE and DeepSeek-V2 — the architecture base DeepSeekMoE (January 2024) introduced the fine-grained expert split that every later DeepSeek MoE model inher
|
||||
its. DeepSeek-V2 (May 2024) then combined that MoE design with Multi-head Latent Attention (MLA) , which is the attention trick that makes long-context serving
|
||||
affordable. DeepSeek-V2 comprises 236B total parameters, of which 21B are activated for each token, and supports a context length of 128K tokens. MLA guarantees
|
||||
efficient inference through significantly compressing the Key-Value (KV) cache into a latent vector, while DeepSeekMoE enables training strong models at an eco
|
||||
nomical cost through sparse computation. Compared with DeepSeek 67B, DeepSeek-V2 achieves significantly stronger performance, and meanwhile saves 42.5% of train
|
||||
ing costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to 5.76 times. Every subsequent paper assumes you know what MLA and fine-
|
||||
grained MoE are. If you read only one paper from 2024, read this one. Our deeper breakdowns sit on the DeepSeek MoE and DeepSeek V2 pages. DeepSeek-Coder and De
|
||||
epSeekMath Two specialist papers from early 2024 matter for the V4 story. DeepSeek-Coder introduced the code-pretraining recipe that fed later general-purpose r
|
||||
eleases. DeepSeekMath introduced Group Relative Policy Optimisation (GRPO) , the RL algorithm that R1 would later use to learn reasoning without human demonstra
|
||||
tions. Readers focused on coding should also see our notes on DeepSeek Coder and the DeepSeek Math page. DeepSeek V3 — the efficiency thesis in paper form The V
|
||||
3 technical report is the one most cited in the January 2025 news cycle. The paper presents DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 6
|
||||
71B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attenti
|
||||
on (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for lo
|
||||
ad balancing and sets a multi-token prediction training objective for stronger performance. DeepSeek-V3 was pre-trained on 14.8 trillion diverse and high-qualit
|
||||
y tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages. The paper’s most-quoted number is the training cost: Combined with 119K GPU hour
|
||||
s for context length extension and 5K GPU hours for post-training, DeepSeek-V3 costs only 2.788M GPU hours for its full training. Assuming the rental price of t
|
||||
he H800 GPU is $2 per GPU hour, our total training costs amount to only $5.576M. That is the number that Reuters and everyone else converted to “about $5.6M.” T
|
||||
he paper is careful to note this excludes prior research and ablations — a caveat many press summaries drop. Our DeepSeek V3 review covers how the claims held u
|
||||
p in production. DeepSeek R1 — reasoning without human demonstrations R1 is the paper that turned DeepSeek from “efficient lab” to “strategic concern” in Wester
|
||||
n press coverage. The core claim is that reasoning steps can be taught with pure RL, no supervised step-by-step reasoning data required. The reasoning abilities
|
||||
of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labeled reasoning trajectories. The proposed RL framework fa
|
||||
cilitates the emergent development of advanced reasoning patterns, such as self-reflection, verification, and dynamic strategy adaptation. The trained model ach
|
||||
ieves superior performance on verifiable tasks such as mathematics, coding competitions, and STEM fields. Two points readers often get wrong: The training cost
|
||||
figure belongs to R1, not V3. Reuters reported DeepSeek’s own R1 training estimate at $294,000 in September 2025. The $5.576M figure in the V3 paper is a separa
|
||||
te model. R1 builds on V3-Base. The paper uses DeepSeek-V3-Base as the base model, DeepSeek-V3 as the instructed model. DeepSeek-R1 and DeepSeek-R1-Zero are tra
|
||||
ined on top of DeepSeek-V3-Base and DeepSeek-R1 reuses non-reasoning data from DeepSeek-V3 SFT data. The two papers are a pair. Our DeepSeek R1 review and the D
|
||||
eepSeek R1 vs OpenAI o1 comparison pair well with the paper. DeepSeek V3.2 — the sparse attention bridge V3.2 is the under-rated paper. Released in December 202
|
||||
5, it is the one that lays the groundwork for V4’s efficiency claims. The paper introduces DeepSeek Sparse Attention (DSA), an efficient attention mechanism tha
|
||||
t substantially reduces computational complexity while preserving model performance in long-context scenarios. By implementing a robust reinforcement learning p
|
||||
rotocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, the high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-
|
||||
5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and
|
||||
the International Collegiate Programming Contest. V3.2’s DSA is instantiated on top of MLA, not as a replacement for it. The V3.2 paper is the most detailed ac
|
||||
count of how DeepSeek thinks about attention efficiency before the bigger leap in V4. The DeepSeek V3.2 page summarises the release notes. DeepSeek V4 — the cur
|
||||
rent paper The V4 technical report, titled DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence , is hosted as a PDF in the V4-Pro repositor
|
||||
y on Hugging Face. It is the primary source for every V4 claim. Architecture contributions The paper’s headline contribution is a new hybrid attention stack. De
|
||||
epSeek-V4 designs a hybrid attention mechanism combining Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to dramatically improve long-c
|
||||
ontext efficiency. In the 1M-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek
|
||||
-V3.2. The paper also incorporates Manifold-Constrained Hyper-Connections (mHC) to strengthen conventional residual connections, enhancing stability of signal p
|
||||
ropagation across layers while preserving model expressivity. Two more items are worth extracting from the report: Training precision. FP4 + FP8 Mixed: MoE expe
|
||||
rt parameters use FP4 precision; most other parameters use FP8. FP4 experts are unusual and are part of why V4 trained at the cost it did. Three reasoning-effor
|
||||
t modes. DeepSeek-V4-Pro and DeepSeek-V4-Flash both support three reasoning effort modes. These map to the API’s non-thinking default, reasoning_effort="high" w
|
||||
ith a thinking flag, and reasoning_effort="max" . In the API this returns reasoning_content alongside the final content . Positioning against closed frontier mo
|
||||
dels The report is unusually candid about where V4 sits. Simon Willison extracted the key quote: “Through the expansion of reasoning tokens, DeepSeek-V4-Pro at
|
||||
maximum reasoning effort demonstrates superior performance relative to GPT-5.2 and Gemini-3.0-Pro on standard reasoning benchmarks. Nevertheless, its performanc
|
||||
e falls marginally short of GPT-5.4 and Gemini-3.1-Pro, suggesting a developmental trajectory that trails current-generation frontier models by approximately 3
|
||||
to 6 months.” That is a self-description the press did not have to spin. On specific benchmarks the paper reports, V4-Pro (reasoning_effort=max) wins on LiveCod
|
||||
eBench (93.5 — a new open-model high), Codeforces rating (3206 — DeepSeek reports this places the model roughly 23rd among human contest participants), Apex Sho
|
||||
rtlist (90.2), SimpleQA-Verified at 57.9 against open and near-open models, and a proof-perfect 120/120 on Putnam-2025 formal reasoning under the hybrid informa
|
||||
l-formal pipeline. Verify any individual number against the PDF before you quote it — frontier benchmark standings shift monthly. How the papers connect: a sugg
|
||||
ested reading order If you have a weekend and want the architecture thread, read in this sequence: DeepSeek-V2 for MLA and DeepSeekMoE — the foundation. DeepSee
|
||||
k-V3 for auxiliary-loss-free load balancing, multi-token prediction, and FP8 training. DeepSeek-R1 for the RL-only reasoning recipe (which uses GRPO from DeepSe
|
||||
ekMath). DeepSeek-V3.2 for DSA as an attention-efficiency prototype. DeepSeek-V4 for the CSA + HCA hybrid and the full mHC / Muon / FP4 training stack. If you o
|
||||
nly have an evening, read the V4 report and V3 report back to back. Together they cover roughly 80% of the current generation’s ideas. The DeepSeek benchmarks 2
|
||||
026 page has the cross-model numbers. Using the papers as a developer, not a reader Every V3 paper onward describes a model you can call through POST /chat/comp
|
||||
letions , the OpenAI-compatible endpoint at https://api.deepseek.com . The current V4 model IDs are deepseek-v4-pro and deepseek-v4-flash . Legacy IDs ( deepsee
|
||||
k-chat , deepseek-reasoner ) remain accepted until 2026-07-24 15:59 UTC, when they retire; both route to deepseek-v4-flash in the meantime. A minimal Python cal
|
||||
l against V4-Pro, matching the paper’s thinking-mode setup, looks like this: from openai import OpenAI client = OpenAI( base_url="https://api.deepseek.com", api
|
||||
_key="YOUR_KEY", ) resp = client.chat.completions.create( model="deepseek-v4-pro", messages=[{"role": "user", "content": "Summarise the V4 paper's DSA section."
|
||||
}], reasoning_effort="high", extra_body={"thinking": {"type": "enabled"}}, ) print(resp.choices[0].message.reasoning_content) print(resp.choices[0].message.cont
|
||||
ent) A few practical details for anyone replicating paper results against the hosted API: The API is stateless. Resend the full messages array with every call.
|
||||
The web chat at chat.deepseek.com keeps session history; the API does not. Context window is 1,000,000 tokens by default on both V4 tiers, with output up to 384
|
||||
,000 tokens. Temperature guidance from DeepSeek’s docs: 0.0 for code and maths (matching the paper’s evaluation setup), 1.0 for data tasks, 1.3 for general chat
|
||||
and translation, 1.5 for creative writing. JSON mode is designed to return valid JSON, not guaranteed. Prompt explicitly with the word “json” plus a small sche
|
||||
ma example, and set max_tokens high enough to avoid truncation. For costs, V4-Flash runs at $0.0028 (cache hit) / $0.14 (cache miss) / $0.28 (output) per 1M tok
|
||||
ens; V4-Pro at promotional $0.435 / $0.87 (75% off through 2026-05-31; list $0.0145 / $1.74 / $3.48). A 1M-call workload with a 2,000-token cached system prompt
|
||||
, a 200-token user message and a 300-token response on V4-Flash works out to $5.60 cached input + $28 uncached input + $84 output = $117.60 total . On V4-Pro th
|
||||
e same workload is $29 + $348 + $1,044 = $1,421 . Check the DeepSeek API pricing page before committing — Preview-window rates can move. For the full API surfac
|
||||
e, start with the DeepSeek API documentation and the DeepSeek API getting started tutorial. What the papers do not tell you Three honest limitations of reading
|
||||
DeepSeek’s papers as your only source: Self-reported benchmarks are self-reported. The V4 report’s evaluation configurations are the paper’s own. Cross-referenc
|
||||
e against independent benchmark aggregators before making a procurement decision. Training-cost figures exclude prior research. The V3 paper says this explicitl
|
||||
y. The $5.576M number is the official run, not the full R&D bill. Licensing is per-model. V4-Pro, V4-Flash, V3.2, V3.1 and R1 publish weights under MIT. Some ol
|
||||
der releases — V3 base, Coder-V2, VL2 — split code (MIT) from weights (a separate DeepSeek Model License). Check the specific Hugging Face repo if licensing mat
|
||||
ters to your deployment. For broader independent context, the DeepSeek latest updates feed tracks post-publication corrections and community replications. DeepS
|
||||
eek AI Guide is an independent resource and is not affiliated with DeepSeek or its parent company. Model IDs, pricing and API behaviour change; check the offici
|
||||
al DeepSeek documentation and pricing page before committing to a production decision. Frequently asked questions Where can I find the official DeepSeek researc
|
||||
h papers? DeepSeek publishes technical reports on arXiv and as PDFs in the matching Hugging Face repository. The V4 report lives inside the V4-Pro model repo; V
|
||||
3 is arXiv 2412.19437; V3.2 is arXiv 2512.02556; R1 is arXiv 2501.12948; V2 is arXiv 2405.04434. A community-maintained index collects the older papers too. For
|
||||
our plain-English summaries of each release, see the DeepSeek models hub . What does the DeepSeek V4 paper actually introduce? The V4 technical report introduc
|
||||
es a hybrid attention stack — Compressed Sparse Attention plus Heavily Compressed Attention — that the authors claim cuts single-token inference FLOPs to 27% an
|
||||
d KV cache to 10% of V3.2 at 1M-token context. It also documents mHC residual connections, Muon-optimised training, and FP4 expert precision. Full details in th
|
||||
e PDF linked from the DeepSeek V4 release date page. How much did it cost to train DeepSeek V3 and R1? The V3 technical report reports 2.788M H800 GPU hours at
|
||||
$2/hour, totalling about $5.576M — explicitly excluding prior research and ablations. DeepSeek’s separately disclosed R1 training cost was around $294,000 (Reut
|
||||
ers, September 2025). The two figures belong to different models and should not be combined. Our DeepSeek V3 review unpacks the cost claim. Is the DeepSeek R1 p
|
||||
aper really a reasoning trace breakthrough? The R1 paper’s contribution is showing that reasoning patterns like self-reflection and verification can emerge from
|
||||
pure reinforcement learning, without human-labelled step-by-step reasoning data. That is a methodological result, not a claim to the best possible reasoning mo
|
||||
del. Compare it honestly against closed competitors in our DeepSeek R1 vs OpenAI o1 analysis. Can I reproduce the benchmark numbers in the papers using the host
|
||||
ed API? Partially. You can match most reasoning and coding benchmarks by setting model="deepseek-v4-pro" , reasoning_effort="max" , temperature=0.0 , and a max_
|
||||
tokens large enough for the reasoning content. Agentic benchmarks depend on tool-use scaffolding that the papers describe but do not ship as a kit. The DeepSeek
|
||||
API getting started tutorial has a minimal client setup. Sources & Methodology Last source check: April 30, 2026 Primary sources Official DeepSeek API document
|
||||
ation Endpoints, request/response shape, model IDs Last checked: April 30, 2026 Pricing DeepSeek API pricing page Per-1M-token rates, billing units, cache-hit p
|
||||
ricing Last checked: April 30, 2026 Model card DeepSeek on Hugging Face Weights, license, usage notes per model Last checked: April 30, 2026 Repository Hugging
|
||||
Face: DeepSeek Papers community collection Community-maintained bibliography of DeepSeek papers Last checked: April 30, 2026 Technical references Technical repo
|
||||
rt DeepSeek-V3 technical report (arXiv:2412.19437) V3 architecture and training methodology Last checked: April 30, 2026 Technical report DeepSeek-R1 paper (arX
|
||||
iv:2501.12948) R1 reasoning approach, distillation results Last checked: April 30, 2026 Technical report DeepSeek LLM scaling-law paper Scaling-law study for 7B
|
||||
and 67B dense models Last checked: April 30, 2026 Technical report DeepSeekMoE paper Fine-grained expert specialisation architecture Last checked: April 30, 20
|
||||
26 Technical report DeepSeek-Coder paper Code-specialised base model recipe Last checked: April 30, 2026 Technical report DeepSeekMath / GRPO paper Group Relati
|
||||
ve Policy Optimisation introduction Last checked: April 30, 2026 Technical report DeepSeek-V2 technical report MLA + MoE architecture foundation for later model
|
||||
s Last checked: April 30, 2026 Technical report DeepSeek-V3 technical report V3 efficiency thesis, training-cost figure Last checked: April 30, 2026 Technical r
|
||||
eport DeepSeek-R1 technical report RL-only reasoning recipe and benchmark results Last checked: April 30, 2026 Technical report DeepSeek-V3.2 technical report (
|
||||
DSA) Sparse attention bridge before V4 Last checked: April 30, 2026 Technical report DeepSeek V4 technical report CSA + HCA hybrid attention, FLOP / KV cache cl
|
||||
aims Last checked: April 30, 2026 Context sources News Reuters: V3 ~$5.6M training cost (Jan 2025) Conversion of GPU-hour figure to dollar headline Last checked
|
||||
: April 30, 2026 News Reuters: R1 $294,000 training cost (Sep 2025) Distinct R1 cost figure separate from V3 Last checked: April 30, 2026 Analysis Simon Williso
|
||||
n: V4 launch-day writeup Extracted key quotes from V4 technical report Last checked: April 30, 2026 Methodology Information in this article was checked against
|
||||
the official DeepSeek documentation, the article's cited primary sources, and the deepseekai.guide pricing dataset on the date shown above. Where third-party nu
|
||||
mbers are quoted, the originating source is linked inline so the reader can verify and re-check. Data confidence High for facts cited inline with primary-source
|
||||
links; medium for context narrative. Editorial note AI tooling moves quickly. If you spot an outdated number or a broken link, the contact link in the footer i
|
||||
s the fastest way to flag it for the editorial team. DG DS Guide Editorial Editorial team DS Guide Editorial is the in-house editorial team at deepseekai.guide.
|
||||
We write practical, source-cited reviews and tutorials for the DeepSeek model family — V4-Pro, V4-Flash, R1, V3, Coder — and the broader frontier-AI ecosystem.
|
||||
Every article carries a Last reviewed date and a Sources section so you can verify what we said and when. Last reviewed: April 30, 2026 · Editorial methodology
|
||||
: view sources Related articles View all in News → News DeepSeek Latest Updates — What Shipped With V4 Preview (April 2026) April 24, 2026 News DeepSeek V4 Rele
|
||||
ase Date: Pro and Flash Land as an Open-Weight Preview April 24, 2026 News DeepSeek Company News: V4, Funding Talks and What Shifted in 2026 April 24, 2026 News
|
||||
DeepSeek Benchmarks 2026: What the V4 Numbers Actually Say April 24, 2026 News DeepSeek Funding News: Inside the $300M First Round April 24, 2026 News DeepSeek
|
||||
Partnerships in the V4 Era: Who’s Actually Shipping April 24, 2026 Leave a Reply Cancel Reply Your email address will not be published. Required fields are mar
|
||||
ked * Name * Email * Website Add Comment * Save my name, email and website in this browser for the next time I comment. Post Comment DeepSeek AI Guide Independe
|
||||
nt reviews, tutorials, and comparisons for the DeepSeek AI ecosystem — no affiliation, no hype. Written and maintained by working AI practitioners. Explore Mode
|
||||
ls Tutorials Comparisons Guides Reviews Alternatives Resources API Docs Free Tools Alternatives News Company About Contact Editorial Guidelines Sitemap Independ
|
||||
ence notice: DeepSeek AI Guide is an independent resource and is not affiliated with, endorsed by, or sponsored by DeepSeek or its parent company. All trademark
|
||||
s and product names are the property of their respective owners and are used for identification purposes only (nominative fair use). © 2026 DeepSeek AI Guide. A
|
||||
ll rights reserved. Privacy Policy Terms of Use Disclaimer Cookie Policy Editorial Guidelines
|
||||
@@ -0,0 +1,148 @@
|
||||
"""Generate the README index for the DeepSeek paper collection.
|
||||
|
||||
Usage: python make_index.py <plan.json> <base_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import sys
|
||||
|
||||
CAT_ORDER = [
|
||||
'01-基础模型', '02-代码模型', '03-数学模型', '04-推理模型', '05-多模态',
|
||||
'06-形式化证明', '07-架构', '08-训练方法', '09-基础设施', '10-其他',
|
||||
]
|
||||
CAT_DESC = {
|
||||
'01-基础模型': '通用大语言模型系列技术报告（LLM / MoE / V2 / V3 / V3.2 / V4 / V4.1-Flash）',
|
||||
'02-代码模型': '代码智能模型与代码驱动推理（DeepSeek-Coder / Coder-V2 / CodeI/O）',
|
||||
'03-数学模型': '数学推理模型（DeepSeekMath / DeepSeekMath-V2）',
|
||||
'04-推理模型': '强化学习驱动的推理模型（DeepSeek-R1）',
|
||||
'05-多模态': '视觉-语言与统一多模态模型（VL / VL2 / Janus / Janus-Pro / JanusFlow / OCR 系列 / DreamCraft3D）',
|
||||
'06-形式化证明': '定理证明与形式化数学（DeepSeek-Prover 系列）',
|
||||
'07-架构': '基础架构创新（mHC 超连接 / Engram 条件记忆 / NSA 原生稀疏注意力 / 无辅助损失负载均衡）',
|
||||
'08-训练方法': '训练与对齐方法（ESFT 专家专精微调 / GRM 通用奖励模型）',
|
||||
'09-基础设施': '训练与推理基础设施（Fire-Flyer AI-HPC / 硬件洞察 / DualPath 存储带宽）',
|
||||
}
|
||||
|
||||
|
||||
def clean_date(p):
|
||||
"""Return a short human date like '11 Feb 2025'."""
|
||||
import re as _re
|
||||
d = p.get('date', '') or ''
|
||||
if d.startswith('[') or len(d) > 24:
|
||||
m = _re.search(r'on\s+(\d{1,2}\s+\w+\s+\d{4})', d)
|
||||
if m:
|
||||
return m.group(1)
|
||||
return d
|
||||
|
||||
|
||||
def main():
|
||||
plan = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
base = sys.argv[2]
|
||||
bycat = {}
|
||||
for p in plan:
|
||||
bycat.setdefault(p['dir'], []).append(p)
|
||||
for c in bycat:
|
||||
bycat[c].sort(key=lambda x: x['stem'])
|
||||
out = []
|
||||
out.append('# DeepSeek 官方论文合集（原文 + 中文翻译）')
|
||||
out.append('')
|
||||
out.append('本目录收录 **深度求索（DeepSeek / DeepSeek-AI）自成立以来以论文形式公开发布的'
|
||||
'全部研究成果**，共 **%d 篇**（截至 2026-09-20）。' % len(plan))
|
||||
out.append('')
|
||||
out.append('每篇论文均包含两部分：')
|
||||
out.append('')
|
||||
out.append('- **原文 PDF**：从 arXiv 下载的官方预印本（未做改动）')
|
||||
out.append('- **中文 Markdown**：摘要全文翻译 + 核心贡献 + 方法要点 + 实验结果 + 结论，'
|
||||
'便于快速掌握论文内容')
|
||||
out.append('')
|
||||
out.append('> 说明：本次中文版本为「摘要 + 核心贡献」精读版。如需某篇论文的全文逐段翻译，'
|
||||
'可在此目录基础上追加。')
|
||||
out.append('')
|
||||
out.append('## 收录范围与判定依据')
|
||||
out.append('')
|
||||
out.append('论文清单通过以下渠道交叉核验后确定，凡判定为 DeepSeek 官方成果者全部收录：')
|
||||
out.append('')
|
||||
out.append('1. **DeepSeek 官网「研究索引」**（https://www.deepseek.com/news ）——官方自行列出的论文条目')
|
||||
out.append('2. **arXiv 作者检索**：作者字段含 `DeepSeek-AI` 的全部论文')
|
||||
out.append('3. **论文署名单位核验**：逐篇抓取 arXiv 摘要页，确认作者单位为 `DeepSeek-AI` / `DeepSeek AI`')
|
||||
out.append('4. **社区索引交叉比对**：Hugging Face `DeepSeek Papers` 合集、第三方综述')
|
||||
out.append('')
|
||||
out.append('判定标准：论文作者单位标注为 DeepSeek-AI / DeepSeek AI，'
|
||||
'或作者主体为 DeepSeek 在职研究员。'
|
||||
'仅「以 DeepSeek 模型为研究对象」的第三方论文（数量众多）**不予收录**。')
|
||||
out.append('')
|
||||
out.append('> **另有一项非论文形式的研究发布**：[LPLB（Linear-Programming-Based Load Balancer）]'
|
||||
'(https://github.com/deepseek-ai/LPLB)——DeepSeek 于 2025 年 11 月 1 日发布的 '
|
||||
'MoE 专家并行负载均衡器，以 GitHub 仓库形式发布（早期研究阶段），'
|
||||
'未上传 arXiv 预印本，故不在本目录论文之列。')
|
||||
out.append('')
|
||||
out.append('## 目录结构')
|
||||
out.append('')
|
||||
out.append('```')
|
||||
out.append('大模型/deepseek/')
|
||||
out.append('├── README.md 本索引')
|
||||
out.append('├── 01-基础模型/ … 09-基础设施/ 各篇论文的原文 PDF + 中文 Markdown（同名配对）')
|
||||
out.append('├── _text/ 论文正文纯文本（供后续「全文逐段翻译」使用）')
|
||||
out.append('├── _digest/ 摘要 + 章节标题 + 正文首尾（精读写作素材）')
|
||||
out.append('└── _scripts/ 清单核验、PDF 下载、正文提取、索引生成脚本')
|
||||
out.append('```')
|
||||
out.append('')
|
||||
out.append('`_text/` 与 `_digest/` 已按 arXiv 编号逐篇存好论文全文，'
|
||||
'后续若要把某几篇扩写为**全文逐段翻译**，可直接基于这些文件进行，无需重新抓取。')
|
||||
out.append('')
|
||||
out.append('> **DreamCraft3D**（2310.16818）为 DeepSeek 参与度有限的早期合作论文：'
|
||||
'作者单位列入 DeepSeek AI，论文注明「实习期间部分完成」，'
|
||||
'为完整起见一并收录并单独标注。')
|
||||
out.append('')
|
||||
out.append('---')
|
||||
out.append('')
|
||||
out.append('## 目录')
|
||||
out.append('')
|
||||
for c in CAT_ORDER:
|
||||
if c in bycat:
|
||||
out.append('- [%s](#%s)（%d 篇）' % (c[3:], c.lower().replace('.', ''), len(bycat[c])))
|
||||
out.append('')
|
||||
out.append('---')
|
||||
out.append('')
|
||||
out.append('## 论文总览')
|
||||
out.append('')
|
||||
out.append('| # | 论文 | arXiv | 发布日期 | 类别 |')
|
||||
out.append('|---|---|---|---|---|')
|
||||
n = 0
|
||||
for c in CAT_ORDER:
|
||||
for p in bycat.get(c, []):
|
||||
n += 1
|
||||
out.append('| %d | [%s](%s) | [%s](%s) | %s | %s |' % (
|
||||
n, p['title'], p['md_path'].replace('\\', '/'), p['id'], p['abs_url'],
|
||||
clean_date(p), c[3:]))
|
||||
out.append('')
|
||||
out.append('---')
|
||||
out.append('')
|
||||
for c in CAT_ORDER:
|
||||
if c not in bycat:
|
||||
continue
|
||||
out.append('## %s' % c[3:])
|
||||
out.append('')
|
||||
out.append('> %s' % CAT_DESC.get(c, ''))
|
||||
out.append('')
|
||||
for p in bycat[c]:
|
||||
stem = p['stem']
|
||||
done = os.path.exists(os.path.join(base, p['md_path']))
|
||||
out.append('### %s' % p['title'])
|
||||
out.append('')
|
||||
out.append('- **arXiv**：%s ｜ **发布**：%s' % (p['id'], clean_date(p)))
|
||||
a = p.get('authors', '')
|
||||
out.append('- **作者**：%s' % (a[:300] + ('…' if len(a) > 300 else '')))
|
||||
out.append('- **原文 PDF**：[%s.pdf](%s)' % (stem, p['pdf_path'].replace('\\', '/')))
|
||||
if done:
|
||||
out.append('- **中文翻译**：[%s.md](%s)' % (stem, p['md_path'].replace('\\', '/')))
|
||||
else:
|
||||
out.append('- **中文翻译**：（待生成）')
|
||||
out.append('')
|
||||
out.append('')
|
||||
with open(os.path.join(base, 'README.md'), 'w', encoding='utf-8') as f:
|
||||
f.write('\n'.join(out))
|
||||
print('README written, papers:', len(plan))
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,45 @@
|
||||
"""Fetch pages and extract arXiv ids + nearby text.
|
||||
|
||||
Usage: python mine_links.py <urls.txt> <out.txt>
|
||||
"""
|
||||
import urllib.request
|
||||
import re
|
||||
import sys
|
||||
import time
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def get(url, timeout=40):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
def strip(s):
|
||||
s = re.sub(r'<script.*?</script>', ' ', s, flags=re.S | re.I)
|
||||
s = re.sub(r'<style.*?</style>', ' ', s, flags=re.S | re.I)
|
||||
s = re.sub(r'<[^>]+>', ' ', s)
|
||||
return ' '.join(htmlmod.unescape(s).split())
|
||||
|
||||
|
||||
urls = [l.strip() for l in open(sys.argv[1], encoding='utf-8') if l.strip()]
|
||||
out = []
|
||||
for u in urls:
|
||||
try:
|
||||
t = get(u)
|
||||
except Exception as e: # noqa: BLE001
|
||||
out.append('=== %s FAILED %s' % (u, str(e)[:100]))
|
||||
continue
|
||||
txt = strip(t)
|
||||
ids = re.findall(r'(?:arxiv[^\s]{0,20}?|arXiv:)(\d{4}\.\d{4,5})', t + ' ' + txt)
|
||||
out.append('=== %s (%d chars, %d arxiv refs)' % (u, len(txt), len(set(ids))))
|
||||
for i in sorted(set(ids)):
|
||||
out.append(' id: ' + i)
|
||||
out.append(' TEXT: ' + txt[:2500])
|
||||
time.sleep(1)
|
||||
|
||||
with open(sys.argv[2], 'w', encoding='utf-8') as f:
|
||||
f.write('\n'.join(out))
|
||||
print('done', len(urls))
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,15 @@
|
||||
=== https://www.deepseek.com/research FAILED HTTP Error 404: Not Found
|
||||
=== https://www.deepseek.com/news (1764 chars, 9 arxiv refs)
|
||||
id: 2505.09343
|
||||
id: 2510.18234
|
||||
id: 2511.22570
|
||||
id: 2512.02556
|
||||
id: 2512.24880
|
||||
id: 2601.07372
|
||||
id: 2601.20552
|
||||
id: 2602.21548
|
||||
id: 2606.19348
|
||||
TEXT: DeepSeek | 研究与动态 获取 APP 扫码下载 DeepSeek App 中文 EN DeepSeek 网页版 DeepSeek 手机应用 API 开放平台 DeepSeek Harness API 文档 加入我们 中文 EN 深度求索，步履不停 投身于探索 AGI 的本质，带着好奇心，用最长期的眼光去回答最大的问题。 动态 2026 年 9 月 10 日 DeepSeek V4.1 Flash：更强、更快、更普惠 DeepSeek V4.1 Flash 正式发布，具备原生多模态视觉理解能力。全新非对称模型结构，能力更强、速度更快、成本更低。 动态 2026 年 4 月 24 日 DeepSeek-V4 预览版：迈入百万上下文普惠时代 DeepSeek 全新系列模型 V4 预览版正式上线并同步开源，拥有百万字超长上下文。 动态 2025 年 12 月 1 日 DeepSeek V3.2 正式版：强化 Agent 能力，融入思考推理 DeepSeek-V3.2 和 V3.2-Speciale 正式发布，推理能力全球领先，首个将思考融入工具使用的模型，斩获 IMO、CMO、ICPC、IOI 金牌。 动态 2025 年 9 月 29 日 DeepSeek-V3.2-Exp 发布，训练推理提效，API 同步降价 DeepSeek-V3.2-Exp 正式发布，引入 DeepSeek Sparse Attention 稀疏注意力机制，API 价格降低 50% 以上。 动态 2025 年 9 月 22 日 DeepSeek-V3.1 版本更新 DeepSeek-V3.1 更新至 Terminus 版本，缓解中英文混杂问题，优化 Code Agent 与 Search Agent 表现。 查看全部 研究索引 2026 年 6 月 24 日 DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence 2026 年 2 月 25 日 DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference 2026 年 1 月 28 日 DeepSeek-OCR 2: Visual Causal Flow 2026 年 1 月 12 日 Engram: Conditional Memory via Scalable Lookup 2025 年 12 月 31 日 mHC: Manifold-Constrained Hyper-Connections 2025 年 12 月 2 日 DeepSeek-V3.2: Pushing the Frontier of Open LLMs 2025 年 11 月 27 日 DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning 2025 年 11 月 1 日 Linear-Programming-Based Load Balancer (LPLB) 2025 年 10 月 21 日 DeepSeek-OCR: Contexts Optical Compression 2025 年 5 月 14 日 Insights into DeepSeek-V3: Scaling Challenges & Hardware Reflections 查看全部 浙ICP备2023025841号 浙B2-20250178 浙公网安备33010502011812号 扫码关注公众号 产品 DeepSeek 网页版 DeepSeek 手机应用 DeepSeek Harness API 开放平台 API 文档 API 价格 服务状态 研究 DeepSeek V4.1 DeepSeek V4 DeepSeek V3.2 DeepSeek V3.1 DeepSeek R1 DeepSeek V3 更多 法务 & 安全 反馈安全漏洞 透明度 加入我们 岗位详情 © 2026 杭州深度求索人工智能基础技术研究有限公司 版权所有 中文 EN
|
||||
=== https://www.deepseek.com/en/research FAILED HTTP Error 404: Not Found
|
||||
=== https://api-docs.deepseek.com/news/news250325 (1737 chars, 0 arxiv refs)
|
||||
TEXT: DeepSeek-V3-0324 Release | DeepSeek API Docs Skip to main content DeepSeek API Docs English English 中文（中国） DeepSeek Platform Quick Start Your First API Call Models & Pricing Token & Token Usage Rate Limit & Isolation Error Codes Agent Integrations API Guides Vision Thinking Mode Multi-round Conversation Chat Prefix Completion (Beta) FIM Completion (Beta) JSON Output Tool Calls Files API Context Caching Using the Responses API Using the Anthropic API API Reference News DeepSeek-V4.1-Flash Release 2026/09/10 DeepSeek-V4-Flash-Vision-Exp Release 2026/08/21 DeepSeek-V4-Pro GA Release 2026/08/13 DeepSeek-V4 Preview Release 2026/04/24 DeepSeek-V3.2 Release 2025/12/01 DeepSeek-V3.2-Exp Release 2025/09/29 DeepSeek V3.1 Update 2025/09/22 DeepSeek V3.1 Release 2025/08/21 DeepSeek-R1-0528 Release 2025/05/28 DeepSeek-V3-0324 Release 2025/03/25 DeepSeek-R1 Release 2025/01/20 DeepSeek APP 2025/01/15 Introducing DeepSeek-V3 2024/12/26 DeepSeek-V2.5-1210 Release 2024/12/10 DeepSeek-R1-Lite Release 2024/11/20 DeepSeek-V2.5 Release 2024/09/05 Context Caching is Available 2024/08/02 New API Features 2024/07/25 Other Resources FAQ Change Log News DeepSeek-V3-0324 Release 2025/03/25 DeepSeek-V3-0324 Release 🔹 Major boost in reasoning performance 🔹 Stronger front-end development skills 🔹 Smarter tool-use capabilities ✅ For non-complex reasoning tasks, we recommend using V3 — just turn off “DeepThink” 🔌 API usage remains unchanged 📜 Models are now released under the MIT License, just like DeepSeek-R1! 🔗 Open-source weights: https://huggingface.co/deepseek-ai/DeepSeek-V3-0324 Previous DeepSeek-R1-0528 Release Next DeepSeek-R1 Release WeChat Official Account Community Email Discord Twitter More GitHub Copyright © 2026 DeepSeek, Inc.
|
||||
@@ -0,0 +1,6 @@
|
||||
=== https://www.deepseek.com/transparency/ (582 chars, 0 arxiv refs)
|
||||
TEXT: DeepSeek | 透明度中心 获取 APP 扫码下载 DeepSeek App 中文 EN DeepSeek 网页版 DeepSeek 手机应用 API 开放平台 DeepSeek Harness API 文档 加入我们 中文 EN 透明度中心 我们在此公开 DeepSeek 已发布的主要模型 模型原理与训练方法说明 查看详情 模型概览 以下为 DeepSeek 已发布的主要模型信息，包含模型名称、发布时间、技术报告与模型卡 V4.0 DeepSeek-V4 New 2026年4月24日 模型卡 技术报告 V3.2 DeepSeek-V3.2 2025年12月1日 模型卡 技术报告 浙ICP备2023025841号 浙B2-20250178 浙公网安备33010502011812号 扫码关注公众号 产品 DeepSeek 网页版 DeepSeek 手机应用 DeepSeek Harness API 开放平台 API 文档 API 价格 服务状态 研究 DeepSeek V4.1 DeepSeek V4 DeepSeek V3.2 DeepSeek V3.1 DeepSeek R1 DeepSeek V3 更多 法务 & 安全 反馈安全漏洞 透明度 加入我们 岗位详情 © 2026 杭州深度求索人工智能基础技术研究有限公司 版权所有 中文 EN
|
||||
=== https://github.com/deepseek-ai/LPLB (7620 chars, 0 arxiv refs)
|
||||
TEXT: GitHub - deepseek-ai/LPLB: An early research stage expert-parallel load balancer for MoE models based on linear programming. · GitHub Skip to content Navigation Menu Sign in Appearance settings Platform AI CODE CREATION GitHub Copilot Write better code with AI GitHub Copilot app Direct agents from issue to merge MCP Registry Integrate external tools DEVELOPER WORKFLOWS Actions Automate any workflow Codespaces Instant dev environments Issues Plan and track work Code Review Manage code changes Code Quality Enforce quality at merge APPLICATION SECURITY GitHub Advanced Security Find and fix vulnerabilities Code security Secure your code as you build Secret protection Stop leaks before they start EXPLORE Why GitHub Documentation Blog Changelog Marketplace View all features Solutions BY COMPANY SIZE Enterprises Small and medium teams Startups Nonprofits BY USE CASE App Modernization DevSecOps DevOps CI/CD View all use cases BY INDUSTRY Healthcare Financial services Manufacturing Government View all industries View all solutions Resources EXPLORE BY TOPIC AI Software Development DevOps Security View all topics EXPLORE BY TYPE Customer stories Events & webinars Ebooks & reports Business insights GitHub Skills SUPPORT & SERVICES Documentation Customer support Community forum Trust center Partners View all resources Open Source COMMUNITY GitHub Sponsors Fund open source developers PROGRAMS Security Lab Maintainer Community GitHub Stars Archive Program REPOSITORIES Topics Trending Collections Enterprise ENTERPRISE SOLUTIONS Enterprise platform AI-powered developer platform AVAILABLE ADD-ONS GitHub Advanced Security Enterprise-grade security features Copilot for Business Enterprise-grade AI features Premium Support Enterprise-grade 24/7 support Pricing Search / Sign in Sign up Appearance settings You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert {{ message }} deepseek-ai / LPLB Public Notifications You must be signed in to change notification settings Fork 43 Star 533 Code Issues 1 Pull requests 1 Discussions Actions Security and quality 0 Insights Additional navigation options Code Issues Pull requests Discussions Actions Security and quality Insights main Branches Tags Go to file Code Open more actions menu Latest commit History 1 Commit 1 Commit Folders and files Name Name Last
|
||||
=== https://github.com/intentlink/deepseek-news-research/blob/main/2025/linear-programming-load-balancer.md (3890 chars, 0 arxiv refs)
|
||||
TEXT: deepseek-news-research/2025/linear-programming-load-balancer.md at main · intentlink/deepseek-news-research · GitHub Skip to content Navigation Menu Sign in Appearance settings Platform AI CODE CREATION GitHub Copilot Write better code with AI GitHub Copilot app Direct agents from issue to merge MCP Registry Integrate external tools DEVELOPER WORKFLOWS Actions Automate any workflow Codespaces Instant dev environments Issues Plan and track work Code Review Manage code changes Code Quality Enforce quality at merge APPLICATION SECURITY GitHub Advanced Security Find and fix vulnerabilities Code security Secure your code as you build Secret protection Stop leaks before they start EXPLORE Why GitHub Documentation Blog Changelog Marketplace View all features Solutions BY COMPANY SIZE Enterprises Small and medium teams Startups Nonprofits BY USE CASE App Modernization DevSecOps DevOps CI/CD View all use cases BY INDUSTRY Healthcare Financial services Manufacturing Government View all industries View all solutions Resources EXPLORE BY TOPIC AI Software Development DevOps Security View all topics EXPLORE BY TYPE Customer stories Events & webinars Ebooks & reports Business insights GitHub Skills SUPPORT & SERVICES Documentation Customer support Community forum Trust center Partners View all resources Open Source COMMUNITY GitHub Sponsors Fund open source developers PROGRAMS Security Lab Maintainer Community GitHub Stars Archive Program REPOSITORIES Topics Trending Collections Enterprise ENTERPRISE SOLUTIONS Enterprise platform AI-powered developer platform AVAILABLE ADD-ONS GitHub Advanced Security Enterprise-grade security features Copilot for Business Enterprise-grade AI features Premium Support Enterprise-grade 24/7 support Pricing Search / Sign in Sign up Appearance settings You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert {{ message }} intentlink / deepseek-news-research Public Notifications You must be signed in to change notification settings Fork 1 Star 17 Code Issues 0 Pull requests 0 Actions Projects Security and quality 0 Insights Additional navigation options Code Issues Pull requests Actions Projects Security and quality Insights Files Expand file tree main Breadcrumbs deepseek-news-research / 2025 / linear-programming-load-balancer.md Copy path Blame More file acti
|
||||
@@ -0,0 +1,35 @@
|
||||
=== https://huggingface.co/collections/Presidentlin/deepseek-papers-674c536aa6acddd9bc98c2ac (7624 chars, 33 arxiv refs)
|
||||
id: 2310.16818
|
||||
id: 2401.02954
|
||||
id: 2401.06066
|
||||
id: 2401.14196
|
||||
id: 2402.03300
|
||||
id: 2403.05525
|
||||
id: 2405.04434
|
||||
id: 2405.14333
|
||||
id: 2406.11931
|
||||
id: 2407.01906
|
||||
id: 2408.08152
|
||||
id: 2408.14158
|
||||
id: 2408.15664
|
||||
id: 2410.13848
|
||||
id: 2411.07975
|
||||
id: 2412.10302
|
||||
id: 2412.19437
|
||||
id: 2501.12948
|
||||
id: 2501.17811
|
||||
id: 2502.07316
|
||||
id: 2502.11089
|
||||
id: 2504.02495
|
||||
id: 2504.21801
|
||||
id: 2505.09343
|
||||
id: 2510.18234
|
||||
id: 2511.22570
|
||||
id: 2512.02556
|
||||
id: 2512.24880
|
||||
id: 2601.07372
|
||||
id: 2601.20552
|
||||
id: 2602.21548
|
||||
id: 2606.19348
|
||||
id: 2609.19969
|
||||
TEXT: Deepseek Papers - a Presidentlin Collection Hugging Face Models Datasets Spaces Buckets new Docs Enterprise Pricing Website Tasks HuggingChat Collections Languages Organizations Community Blog Posts Daily Papers Hardware Learn Discord Forum GitHub Solutions Team & Enterprise Hugging Face PRO Enterprise Support Inference Providers Inference Endpoints Storage Buckets Log In Sign Up Presidentlin 's Collections ByteDance Papers Deepseek Papers AI Release Year Thread 2026 AI Release Year Thread 2025 AI Release Week Thread (07 Sept 2026) AI Release Week Thread (31 August 2026) AI Release Week Thread (24 August 2026) AI Release Week Thread (17 August 2026) AI Release Week Thread (10 August 2026) AI Release Week Thread (03 August 2026) AI Release Week Thread (27 July 2026) AI Release Week Thread (20 July 2026) AI Release Week Thread (13 July 2026) AI Release Week Thread (6 July 2026) AI Release Week Thread (29 June 2026) AI Release Week Thread (22 June 2026) AI Release Week Thread (15 June 2026) AI Release Week Thread (08 June 2026) AI Release Week Thread (01 June 2026) AI Release Week Thread (25 May 2026) AI Release Week Thread (18 May 2026) AI Release Week Thread (11 May 2026) AI Release Week Thread (04 May 2026) AI Release Week Thread (27 April 2026) AI Release Week Thread (20 April 2026) AI Release Week Thread (13 April 2026) AI Release Week Thread (6 April 2026) AI Release Week Thread (30 March 2026) AI Release Week Thread (23 March 2026) AI Release Week Thread (16 March 2026) AI Release Week Thread (09 March 2026) AI Release Week Thread (02 March 2026) AI Release Week Thread (23 February 2026) AI Release Week Thread (16 February 2026) AI Release Week Thread (09 February 2026) AI Release Week Thread (02 February 2026) AI Release Week Thread (26 January 2026) AI Release Week Thread (19 January 2026) AI Release Week Thread (12 January 2026) AI Release Week Thread (05 January 2026) AI Release Week Thread (29 December 2025) AI Release Week Thread (22 December 2025) AI Release Week Thread (15 December 2025) AI Release Week Thread (8 December 2025) AI Release Week Thread (1 December 2025) AI Release Week Thread (24 November 2025) AI Release Week Thread (10 November 2025) AI Release Week Thread (3 November 2025) AI Release Week Thread (27 October 2025) AI Release Week Thread (20 October 2025) AI Release Week Thread (13 October 2025) AI Release Week Thread (6 October 2025) AI Release Week Thread (29 September 2025) AI Release Week Thread (22 September 2025) AI Releas
|
||||
@@ -0,0 +1,89 @@
|
||||
"""Organize PDFs into category folders and emit a translation work plan.
|
||||
|
||||
Usage: python organize.py <papers.json> <base_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import shutil
|
||||
import sys
|
||||
|
||||
CAT_DIRS = {
|
||||
'基础模型': '01-基础模型',
|
||||
'代码模型': '02-代码模型',
|
||||
'数学模型': '03-数学模型',
|
||||
'推理模型': '04-推理模型',
|
||||
'多模态': '05-多模态',
|
||||
'形式化证明': '06-形式化证明',
|
||||
'架构': '07-架构',
|
||||
'训练方法': '08-训练方法',
|
||||
'基础设施': '09-基础设施',
|
||||
'其他': '10-其他',
|
||||
}
|
||||
|
||||
SHORT = {
|
||||
'2401.02954': 'DeepSeek-LLM',
|
||||
'2401.06066': 'DeepSeekMoE',
|
||||
'2401.14196': 'DeepSeek-Coder',
|
||||
'2402.03300': 'DeepSeekMath',
|
||||
'2403.05525': 'DeepSeek-VL',
|
||||
'2405.04434': 'DeepSeek-V2',
|
||||
'2405.14333': 'DeepSeek-Prover',
|
||||
'2406.11931': 'DeepSeek-Coder-V2',
|
||||
'2407.01906': 'ESFT-专家专精微调',
|
||||
'2408.08152': 'DeepSeek-Prover-V1.5',
|
||||
'2408.14158': 'Fire-Flyer-AI-HPC',
|
||||
'2410.13848': 'Janus',
|
||||
'2412.10302': 'DeepSeek-VL2',
|
||||
'2412.19437': 'DeepSeek-V3',
|
||||
'2501.12948': 'DeepSeek-R1',
|
||||
'2501.17811': 'Janus-Pro',
|
||||
'2504.02495': 'DeepSeek-GRM',
|
||||
'2504.21801': 'DeepSeek-Prover-V2',
|
||||
'2505.09343': 'Insights-into-DeepSeek-V3',
|
||||
'2510.18234': 'DeepSeek-OCR',
|
||||
'2511.22570': 'DeepSeekMath-V2',
|
||||
'2512.02556': 'DeepSeek-V3.2',
|
||||
'2512.24880': 'mHC',
|
||||
'2601.07372': 'Engram-条件记忆',
|
||||
'2601.20552': 'DeepSeek-OCR-2',
|
||||
'2602.21548': 'DualPath',
|
||||
'2606.19348': 'DeepSeek-V4',
|
||||
'2609.19969': 'DeepSeek-V4.1-Flash',
|
||||
}
|
||||
|
||||
|
||||
def main():
|
||||
papers = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
base = sys.argv[2]
|
||||
pdfdir = os.path.join(base, 'pdf')
|
||||
plan = []
|
||||
for cat in CAT_DIRS:
|
||||
pass
|
||||
counters = {}
|
||||
for p in papers:
|
||||
cat = p.get('category', '其他')
|
||||
d = CAT_DIRS.get(cat, '10-其他')
|
||||
counters[d] = counters.get(d, 0) + 1
|
||||
nn = '%02d' % counters[d]
|
||||
short = SHORT.get(p['id'], p['id'])
|
||||
stem = '%s-%s-%s' % (nn, short, p['id'])
|
||||
outdir = os.path.join(base, d)
|
||||
os.makedirs(outdir, exist_ok=True)
|
||||
src = os.path.join(pdfdir, p['id'] + '.pdf')
|
||||
dst = os.path.join(outdir, stem + '.pdf')
|
||||
if os.path.exists(src) and not os.path.exists(dst):
|
||||
shutil.move(src, dst)
|
||||
p['dir'] = d
|
||||
p['stem'] = stem
|
||||
p['pdf_path'] = os.path.join(d, stem + '.pdf')
|
||||
p['md_path'] = os.path.join(d, stem + '.md')
|
||||
plan.append(p)
|
||||
with open(os.path.join(base, '_scripts', 'plan.json'), 'w', encoding='utf-8') as f:
|
||||
json.dump(plan, f, ensure_ascii=False, indent=1)
|
||||
for p in plan:
|
||||
print(p['dir'], '|', p['stem'], '|', p['title'][:60])
|
||||
print('total', len(plan))
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,20 @@
|
||||
"""Dump full page text to a file.
|
||||
|
||||
Usage: python page_text.py <url> <out.txt>
|
||||
"""
|
||||
import urllib.request
|
||||
import re
|
||||
import sys
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
req = urllib.request.Request(sys.argv[1], headers={'User-Agent': UA})
|
||||
t = urllib.request.urlopen(req, timeout=45).read().decode('utf-8', 'ignore')
|
||||
t = re.sub(r'<script.*?</script>', ' ', t, flags=re.S | re.I)
|
||||
t = re.sub(r'<style.*?</style>', ' ', t, flags=re.S | re.I)
|
||||
t = re.sub(r'<[^>]+>', ' ', t)
|
||||
t = ' '.join(htmlmod.unescape(t).split())
|
||||
with open(sys.argv[2], 'w', encoding='utf-8') as f:
|
||||
for i in range(0, len(t), 160):
|
||||
f.write(t[i:i + 160] + '\n')
|
||||
print('chars', len(t))
|
||||
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,87 @@
|
||||
"""Scrape arXiv advanced search by author name, all pages.
|
||||
|
||||
Usage: python scrape_author.py <author> <out.json>
|
||||
"""
|
||||
import urllib.request
|
||||
import urllib.parse
|
||||
import re
|
||||
import json
|
||||
import sys
|
||||
import time
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def get(url, timeout=45):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
def strip(s):
|
||||
s = re.sub(r'<[^>]+>', ' ', s)
|
||||
return ' '.join(htmlmod.unescape(s).split())
|
||||
|
||||
|
||||
def parse_page(t):
|
||||
out = []
|
||||
for blk in re.split(r'<li class="arxiv-result">', t)[1:]:
|
||||
aid = re.search(r'arxiv\.org/abs/([^"\'<]+)', blk)
|
||||
title = re.search(r'<p class="title is-5 mathjax">(.*?)</p>', blk, re.S)
|
||||
authors = re.search(r'<p class="authors">(.*?)</p>', blk, re.S)
|
||||
dates = re.search(r'<p class="is-size-7">(.*?)</p>', blk, re.S)
|
||||
if not aid:
|
||||
continue
|
||||
out.append({
|
||||
'id': aid.group(1),
|
||||
'title': strip(title.group(1)) if title else '',
|
||||
'authors': strip(authors.group(1)) if authors else '',
|
||||
'dates': strip(dates.group(1)) if dates else '',
|
||||
})
|
||||
return out
|
||||
|
||||
|
||||
def main():
|
||||
term = sys.argv[1]
|
||||
out_path = sys.argv[2]
|
||||
size = 200
|
||||
res = []
|
||||
page = 0
|
||||
while True:
|
||||
qs = [
|
||||
('advanced', ''),
|
||||
('terms-0-operator', 'AND'),
|
||||
('terms-0-term', term),
|
||||
('terms-0-field', 'author'),
|
||||
('classification-include_cross_list', 'include'),
|
||||
('start', str(page * size)),
|
||||
('size', str(size)),
|
||||
('order', '-announced_date_first'),
|
||||
]
|
||||
url = 'https://arxiv.org/search/advanced?' + urllib.parse.urlencode(qs)
|
||||
try:
|
||||
t = get(url)
|
||||
except Exception as e: # noqa: BLE001
|
||||
print('page fail', page, e, file=sys.stderr)
|
||||
break
|
||||
rows = parse_page(t)
|
||||
total = re.search(r'of\s+([\d,]+)\s+results', t)
|
||||
print('page', page, 'rows', len(rows), 'total', total.group(1) if total else '?', file=sys.stderr)
|
||||
res.extend(rows)
|
||||
if len(rows) < size:
|
||||
break
|
||||
page += 1
|
||||
time.sleep(2)
|
||||
if page > 20:
|
||||
break
|
||||
seen = {}
|
||||
for r in res:
|
||||
seen.setdefault(r['id'].split('v')[0], r)
|
||||
with open(out_path, 'w', encoding='utf-8') as f:
|
||||
json.dump(list(seen.values()), f, ensure_ascii=False, indent=1)
|
||||
print('saved', len(seen), file=sys.stderr)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,78 @@
|
||||
"""Scrape arXiv search HTML for a query and dump id/title/authors.
|
||||
|
||||
Usage: python scrape_search.py <query> <pages> <out.json>
|
||||
"""
|
||||
import urllib.request
|
||||
import urllib.parse
|
||||
import re
|
||||
import json
|
||||
import sys
|
||||
import time
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def get(url, timeout=40):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
def strip(s):
|
||||
s = re.sub(r'<[^>]+>', ' ', s)
|
||||
s = htmlmod.unescape(s)
|
||||
return ' '.join(s.split())
|
||||
|
||||
|
||||
def parse_page(t):
|
||||
out = []
|
||||
# each result block
|
||||
for m in re.finditer(r'<li class="arxiv-result">(.*?)</li>\s*(?=<li class="arxiv-result">|</ol>)', t, re.S):
|
||||
blk = m.group(1)
|
||||
aid = re.search(r'arxiv\.org/abs/([^"\'<]+)', blk)
|
||||
title = re.search(r'<p class="title is-5 mathjax">(.*?)</p>', blk, re.S)
|
||||
authors = re.search(r'<p class="authors">(.*?)</p>', blk, re.S)
|
||||
dates = re.search(r'<p class="is-size-7">(.*?)</p>', blk, re.S)
|
||||
if not aid:
|
||||
continue
|
||||
out.append({
|
||||
'id': aid.group(1),
|
||||
'title': strip(title.group(1)) if title else '',
|
||||
'authors': strip(authors.group(1)) if authors else '',
|
||||
'dates': strip(dates.group(1)) if dates else '',
|
||||
})
|
||||
return out
|
||||
|
||||
|
||||
def main():
|
||||
q = sys.argv[1]
|
||||
pages = int(sys.argv[2]) if len(sys.argv) > 2 else 1
|
||||
out_path = sys.argv[3] if len(sys.argv) > 3 else 'search.json'
|
||||
size = 200
|
||||
res = []
|
||||
for p in range(pages):
|
||||
url = ('https://arxiv.org/search/?searchtype=all&query=' + urllib.parse.quote(q)
|
||||
+ '&start=' + str(p * size) + '&size=' + str(size))
|
||||
try:
|
||||
t = get(url)
|
||||
except Exception as e: # noqa: BLE001
|
||||
print('page fail', p, e, file=sys.stderr)
|
||||
break
|
||||
rows = parse_page(t)
|
||||
total = re.search(r'of\s+([\d,]+)\s+results', t)
|
||||
print('page', p, 'rows', len(rows), 'total', total.group(1) if total else '?', file=sys.stderr)
|
||||
res.extend(rows)
|
||||
if len(rows) < size:
|
||||
break
|
||||
time.sleep(2)
|
||||
seen = {}
|
||||
for r in res:
|
||||
seen.setdefault(r['id'].split('v')[0], r)
|
||||
with open(out_path, 'w', encoding='utf-8') as f:
|
||||
json.dump(list(seen.values()), f, ensure_ascii=False, indent=1)
|
||||
print('saved', len(seen), file=sys.stderr)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,89 @@
|
||||
"""Scrape arXiv advanced search (title contains term), all pages.
|
||||
|
||||
Usage: python scrape_title.py <term> <out.json>
|
||||
"""
|
||||
import urllib.request
|
||||
import urllib.parse
|
||||
import re
|
||||
import json
|
||||
import sys
|
||||
import time
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def get(url, timeout=45):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
def strip(s):
|
||||
s = re.sub(r'<[^>]+>', ' ', s)
|
||||
return ' '.join(htmlmod.unescape(s).split())
|
||||
|
||||
|
||||
def parse_page(t):
|
||||
out = []
|
||||
blocks = re.split(r'<li class="arxiv-result">', t)[1:]
|
||||
for blk in blocks:
|
||||
aid = re.search(r'arxiv\.org/abs/([^"\'<]+)', blk)
|
||||
title = re.search(r'<p class="title is-5 mathjax">(.*?)</p>', blk, re.S)
|
||||
authors = re.search(r'<p class="authors">(.*?)</p>', blk, re.S)
|
||||
dates = re.search(r'<p class="is-size-7">(.*?)</p>', blk, re.S)
|
||||
if not aid:
|
||||
continue
|
||||
out.append({
|
||||
'id': aid.group(1),
|
||||
'title': strip(title.group(1)) if title else '',
|
||||
'authors': strip(authors.group(1)) if authors else '',
|
||||
'dates': strip(dates.group(1)) if dates else '',
|
||||
})
|
||||
return out
|
||||
|
||||
|
||||
def main():
|
||||
term = sys.argv[1]
|
||||
out_path = sys.argv[2]
|
||||
size = 200
|
||||
res = []
|
||||
page = 0
|
||||
while True:
|
||||
qs = [
|
||||
('advanced', ''),
|
||||
('terms-0-operator', 'AND'),
|
||||
('terms-0-term', term),
|
||||
('terms-0-field', 'title'),
|
||||
('classification-physics_archives', 'all'),
|
||||
('classification-include_cross_list', 'include'),
|
||||
('start', str(page * size)),
|
||||
('size', str(size)),
|
||||
('order', '-announced_date_first'),
|
||||
]
|
||||
url = 'https://arxiv.org/search/advanced?' + urllib.parse.urlencode(qs)
|
||||
try:
|
||||
t = get(url)
|
||||
except Exception as e: # noqa: BLE001
|
||||
print('page fail', page, e, file=sys.stderr)
|
||||
break
|
||||
rows = parse_page(t)
|
||||
total = re.search(r'of\s+([\d,]+)\s+results', t)
|
||||
print('page', page, 'rows', len(rows), 'total', total.group(1) if total else '?', file=sys.stderr)
|
||||
res.extend(rows)
|
||||
if len(rows) < size:
|
||||
break
|
||||
page += 1
|
||||
time.sleep(2)
|
||||
if page > 15:
|
||||
break
|
||||
seen = {}
|
||||
for r in res:
|
||||
seen.setdefault(r['id'].split('v')[0], r)
|
||||
with open(out_path, 'w', encoding='utf-8') as f:
|
||||
json.dump(list(seen.values()), f, ensure_ascii=False, indent=1)
|
||||
print('saved', len(seen), file=sys.stderr)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,13 @@
|
||||
import json
|
||||
import sys
|
||||
|
||||
p = sys.argv[1] if len(sys.argv) > 1 else 'verified.json'
|
||||
with open(p, encoding='utf-8') as f:
|
||||
rows = json.load(f)
|
||||
for r in rows:
|
||||
a = r.get('authors', '')
|
||||
print('=' * 100)
|
||||
print(r['id'], '|', r.get('dateline', ''))
|
||||
print('T:', r.get('title', ''))
|
||||
print('A:', a[:320])
|
||||
print('C:', r.get('comments', '')[:200])
|
||||
@@ -0,0 +1,48 @@
|
||||
"""Per-paper completion report.
|
||||
|
||||
Usage: python status_report.py <plan.json> <base_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
|
||||
plan = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
base = sys.argv[2]
|
||||
SECS = ['一、', '二、', '三、', '四、', '五、', '六、']
|
||||
|
||||
rows = []
|
||||
incomplete = []
|
||||
for p in plan:
|
||||
pdf = os.path.join(base, p['pdf_path'])
|
||||
md = os.path.join(base, p['md_path'])
|
||||
pdf_ok = os.path.exists(pdf) and open(pdf, 'rb').read(4) == b'%PDF' and os.path.getsize(pdf) > 20000
|
||||
md_ok = os.path.exists(md)
|
||||
cjk = 0
|
||||
miss = []
|
||||
link_ok = False
|
||||
if md_ok:
|
||||
t = open(md, encoding='utf-8').read()
|
||||
cjk = len(re.findall(r'[\u4e00-\u9fff]', t))
|
||||
for s in SECS:
|
||||
if '## ' + s not in t:
|
||||
miss.append(s[0])
|
||||
m = re.search(r'\]\(\./([^)]+\.pdf)\)', t)
|
||||
if m:
|
||||
link_ok = os.path.exists(os.path.join(os.path.dirname(md), m.group(1)))
|
||||
if md_ok and not miss and link_ok and pdf_ok:
|
||||
pass
|
||||
else:
|
||||
incomplete.append((p['id'], p['stem'], pdf_ok, md_ok, miss, link_ok))
|
||||
rows.append((p['dir'], p['stem'], pdf_ok, md_ok, cjk, ''.join(miss) or '-', link_ok))
|
||||
|
||||
print('%-16s %-46s %-6s %-6s %-7s %-8s %s' % ('类别', '文件前缀', 'PDF', 'MD', '汉字', '缺章节', 'PDF链接'))
|
||||
print('-' * 120)
|
||||
for d, stem, po, mo, cjk, miss, lo in rows:
|
||||
print('%-16s %-46s %-6s %-6s %-7d %-8s %s' % (
|
||||
d, stem, 'OK' if po else 'MISSING', 'OK' if mo else 'MISSING',
|
||||
cjk, miss, 'OK' if lo else 'BROKEN'))
|
||||
print('-' * 120)
|
||||
print('总计 %d 篇；未完成 %d 篇' % (len(rows), len(incomplete)))
|
||||
for i in incomplete:
|
||||
print(' ! 未完成:', i)
|
||||
@@ -0,0 +1,770 @@
|
||||
[
|
||||
{
|
||||
"id": "2609.19969",
|
||||
"title": "DeepSeek -V4.1-Flash: Pushing the Limits of KV Cache Compression",
|
||||
"authors": "Authors: DeepSeek-AI , : , Anyi Xu , B. Li , Bangcai Lin , Bing Xue , BingCheng Xian , Bingzheng Xu , Bochao Wu , Bowei Zhang , Boyi Deng , C. C. Yu , Chao Jin , Chaofan Lin , Chen Dong , Chenbing Wang , Chenfan Feng , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chengyuan Zhang , Chenhao Xu , Chenqi Zhao , Chenze Shao , Chuhao Wang , et al. (568 additional authors not shown)",
|
||||
"dates": "Submitted 17 September, 2026; originally announced September 2026."
|
||||
},
|
||||
{
|
||||
"id": "2609.18038",
|
||||
"title": "CoAtNet-DeepMoE: A Convolution-Attention Hybrid with DeepSeek Mixture-of-Experts for Parameter-Efficient Tomato Disease Classification",
|
||||
"authors": "Authors: Md Nadim Mahamood , Md Arif Shahriar , Md Shafi Ud Doula , Kamrul Hasan",
|
||||
"dates": "Submitted 15 September, 2026; originally announced September 2026."
|
||||
},
|
||||
{
|
||||
"id": "2609.15627",
|
||||
"title": "DeepSeek -V4-Flash on AMD gfx90a: Correctness Recovery and Inference Performance Engineering",
|
||||
"authors": "Authors: Siming Huang",
|
||||
"dates": "Submitted 14 September, 2026; originally announced September 2026."
|
||||
},
|
||||
{
|
||||
"id": "2609.07008",
|
||||
"title": "RedKnot-MLA: Multi-Head Offline-Online Reuse for DeepSeek -V4 Long-Context Serving",
|
||||
"authors": "Authors: Yang Liu , Zhaokai Luo , Huayi Jin , Ruozhou He , Chenchen Hong , Mingxiao Ma , Biao Zhang , Zhiyong Wang , Boyu Wang , Guanjie Chen , Yifei Liu , Tao Xie , Junhao Hu",
|
||||
"dates": "Submitted 6 September, 2026; originally announced September 2026."
|
||||
},
|
||||
{
|
||||
"id": "2608.30877",
|
||||
"title": "Deploying DeepSeek 175B Locally on a Single Consumer-Grade RTX 4060 Laptop with 32GB RAM for 200k-Scale Protein-Ligand Virtual Screening",
|
||||
"authors": "Authors: Rui Xiao , Yili Xu",
|
||||
"dates": "Submitted 31 August, 2026; originally announced August 2026."
|
||||
},
|
||||
{
|
||||
"id": "2608.16393",
|
||||
"title": "Security Assessment of DeepSeek Harness with A.I.G: Evaluating Resistance to Indirect Prompt Injection",
|
||||
"authors": "Authors: Zonghao Ying , Xiangfan Wu , Huiyu Wu , Xing Zheng , Huangsheng Cheng , Xiaorong Shi , Jing Guo",
|
||||
"dates": "Submitted 18 August, 2026; v1 submitted 17 August, 2026; originally announced August 2026."
|
||||
},
|
||||
{
|
||||
"id": "2607.20145",
|
||||
"title": "SLAI T-Rex: Full-Parameter Post-training of the DeepSeek -V4 Family on Ascend SuperPOD",
|
||||
"authors": "Authors: Dongfang Li , Xiaodong Luo , Ruoyu Sun , Xuhui Chen , Linyuan Qiu , Jian Meng , Zhengxuan Lu , Yiting Wang , Yucheng Xie , Tao Guo , Tianxiang Fang , Jing Li , Sihang Chen , Shihao Hong , Chang Liu , Weihua Dai , Zirong Zeng , Ziwei Zhu , Zhuohan Wang , Zhengjun Yue , Igor Vasilyev , Min Liu , Weijian Sun , Xin Chen , Yingmeng Gao , et al. (40 additional authors not shown)",
|
||||
"dates": "Submitted 19 August, 2026; v1 submitted 22 July, 2026; originally announced July 2026."
|
||||
},
|
||||
{
|
||||
"id": "2606.19348",
|
||||
"title": "DeepSeek -V4: Towards Highly Efficient Million-Token Context Intelligence",
|
||||
"authors": "Authors: DeepSeek-AI , Anyi Xu , Bangcai Lin , Bing Xue , Bingxuan Wang , Bingzheng Xu , Bochao Wu , Bowei Zhang , Chaofan Lin , Chen Dong , Chenchen Ling , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chengyu Hou , Chenhao Xu , Chenze Shao , Chong Ruan , Conner Sun , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Donghao Li , Dongjie Ji , et al. (294 additional authors not shown)",
|
||||
"dates": "Submitted 26 April, 2026; originally announced June 2026."
|
||||
},
|
||||
{
|
||||
"id": "2606.10392",
|
||||
"title": "Instruction Finetuning DeepSeek -R1-8B Model Using LoRA and NEFTune",
|
||||
"authors": "Authors: Wu Yuerong , Mingni Luo",
|
||||
"dates": "Submitted 9 June, 2026; originally announced June 2026."
|
||||
},
|
||||
{
|
||||
"id": "2606.09079",
|
||||
"title": "FlashMemory- DeepSeek -V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention",
|
||||
"authors": "Authors: Yan Wang , Qifan Zhang , Jiachen Yu , Tian Liang , Dongyang Ma , Xiang Hu , Zibo Lin , Chunyang Li , Zhichao Wang , Miao Peng , Nuo Chen , Jia Li , Yujiu Yang , Haitao Mi , Dong Yu",
|
||||
"dates": "Submitted 20 July, 2026; v1 submitted 8 June, 2026; originally announced June 2026."
|
||||
},
|
||||
{
|
||||
"id": "2606.07410",
|
||||
"title": "A Comprehensive Anatomy of Human and DeepSeek -R1 LLM Mathematical Reasoning",
|
||||
"authors": "Authors: Yuxiang Chen , Jun Wang",
|
||||
"dates": "Submitted 5 June, 2026; originally announced June 2026."
|
||||
},
|
||||
{
|
||||
"id": "2606.01228",
|
||||
"title": "Institutional Trust and the Domestic AI Advantage: Evidence from DeepSeek and ChatGPT Users in China",
|
||||
"authors": "Authors: Jiashen Huang , Yu Jia , Xu Pan",
|
||||
"dates": "Submitted 31 May, 2026; originally announced June 2026."
|
||||
},
|
||||
{
|
||||
"id": "2605.10234",
|
||||
"title": "Social Policy of Large Language Models: How GPT, Claude, DeepSeek and Grok Allocate Social Budgets in Spain and Germany",
|
||||
"authors": "Authors: Claudia Benavides Cantos , Eduardo C. Garrido-Merchán",
|
||||
"dates": "Submitted 11 May, 2026; originally announced May 2026."
|
||||
},
|
||||
{
|
||||
"id": "2605.08615",
|
||||
"title": "DSPE: An Energy-Efficient Edge Processor for DeepSeek Inference with MerkleTree-based Incremental Pruning, Multi-Stage Boothing Lookup and Dynamic Adaptive Posit Processing",
|
||||
"authors": "Authors: Yuhan Zhang , Zhou Wang , Zhou Shu , Jiuren Zhou , Yanqing Xu , Xiaonan Tang , Shushan Qiao , Tianchun Ye , Yang Liu , Anil A. Bharath , Emm Mic Drakakis",
|
||||
"dates": "Submitted 8 May, 2026; originally announced May 2026."
|
||||
},
|
||||
{
|
||||
"id": "2605.00392",
|
||||
"title": "RTPrune: Reading-Twice Inspired Token Pruning for Efficient DeepSeek -OCR Inference",
|
||||
"authors": "Authors: Ben Wan , Yan Feng , Zihan Tang , Weizhe Huang , Yuting Zeng , Jia Wang , Tongxuan Liu",
|
||||
"dates": "Submitted 21 May, 2026; v1 submitted 1 May, 2026; originally announced May 2026."
|
||||
},
|
||||
{
|
||||
"id": "2604.12548",
|
||||
"title": "DeepSeek Robustness Against Semantic-Character Dual-Space Mutated Prompt Injection",
|
||||
"authors": "Authors: Junyu Ren , Xingjian Pan , Wensheng Gan , Philip S. Yu",
|
||||
"dates": "Submitted 14 April, 2026; originally announced April 2026."
|
||||
},
|
||||
{
|
||||
"id": "2604.03476",
|
||||
"title": "Fine-tuning DeepSeek -OCR-2 for Molecular Structure Recognition",
|
||||
"authors": "Authors: Haocheng Tang , Xingyu Dang , Junmei Wang",
|
||||
"dates": "Submitted 20 April, 2026; v1 submitted 3 April, 2026; originally announced April 2026."
|
||||
},
|
||||
{
|
||||
"id": "2603.26673",
|
||||
"title": "Can AI be a Teaching Partner? Evaluating ChatGPT, Gemini, and DeepSeek across Three Teaching Strategies",
|
||||
"authors": "Authors: Talita de Paula Cypriano de Souza , Shruti Mehta , Matheus Arataque Uema , Luciano Bernardes de Paula , Seiji Isotani",
|
||||
"dates": "Submitted 23 February, 2026; originally announced March 2026."
|
||||
},
|
||||
{
|
||||
"id": "2603.19253",
|
||||
"title": "A comprehensive study of LLM-based argument classification: from Llama through DeepSeek to GPT-5.2",
|
||||
"authors": "Authors: Marcin Pietroń , Filip Gampel , Jakub Gomułka , Andrzej Tomski , Rafał Olszowski",
|
||||
"dates": "Submitted 25 February, 2026; originally announced March 2026."
|
||||
},
|
||||
{
|
||||
"id": "2602.14158",
|
||||
"title": "A Multi-Agent Framework for Medical AI: Leveraging Fine-Tuned GPT, LLaMA, and DeepSeek R1 for Evidence-Based and Bias-Aware Clinical Query Processing",
|
||||
"authors": "Authors: Naeimeh Nourmohammadi , Md Meem Hossain , The Anh Han , Safina Showkat Ara , Zia Ush Shamszaman",
|
||||
"dates": "Submitted 15 February, 2026; originally announced February 2026."
|
||||
},
|
||||
{
|
||||
"id": "2601.20552",
|
||||
"title": "DeepSeek -OCR 2: Visual Causal Flow",
|
||||
"authors": "Authors: Haoran Wei , Yaofeng Sun , Yukun Li",
|
||||
"dates": "Submitted 28 January, 2026; originally announced January 2026."
|
||||
},
|
||||
{
|
||||
"id": "2601.11464",
|
||||
"title": "MHA2MLA-VLM: Enabling DeepSeek's Economical Multi-Head Latent Attention across Vision-Language Models",
|
||||
"authors": "Authors: Xiaoran Fan , Zhichao Sun , Tao Ji , Lixing Shen , Tao Gui",
|
||||
"dates": "Submitted 16 January, 2026; originally announced January 2026."
|
||||
},
|
||||
{
|
||||
"id": "2601.03714",
|
||||
"title": "Visual Merit or Linguistic Crutch? A Close Look at DeepSeek -OCR",
|
||||
"authors": "Authors: Yunhao Liang , Ruixuan Ying , Bo Li , Hong Li , Kai Yan , Qingwen Li , Min Yang , Okamoto Satoshi , Zhe Cui , Shiwen Ni",
|
||||
"dates": "Submitted 8 January, 2026; v1 submitted 7 January, 2026; originally announced January 2026."
|
||||
},
|
||||
{
|
||||
"id": "2512.20344",
|
||||
"title": "A DeepSeek -Powered AI System for Automated Chest Radiograph Interpretation in Clinical Practice",
|
||||
"authors": "Authors: Yaowei Bai , Ruiheng Zhang , Yu Lei , Xuhua Duan , Jingfeng Yao , Shuguang Ju , Chaoyang Wang , Wei Yao , Yiwan Guo , Guilin Zhang , Chao Wan , Qian Yuan , Lei Chen , Wenjuan Tang , Biqiang Zhu , Xinggang Wang , Tao Sun , Wei Zhou , Dacheng Tao , Yongchao Xu , Chuansheng Zheng , Huangxuan Zhao , Bo Du",
|
||||
"dates": "Submitted 23 December, 2025; originally announced December 2025."
|
||||
},
|
||||
{
|
||||
"id": "2512.10576",
|
||||
"title": "ESS: An Offload-Centric Latent-Cache Management Architecture for DeepSeek -V3.2-Exp",
|
||||
"authors": "Authors: Xinhang Chen , Chao Zhang , Jiahuan He , Wei Liu , Jianming Zhang , Wenlong Zhou , Xiao Li , Pai Zeng , Shiyong Li , Yuanpan Qian , Dong Li , Zhaogeng Li",
|
||||
"dates": "Submitted 11 December, 2025; originally announced December 2025."
|
||||
},
|
||||
{
|
||||
"id": "2512.09772",
|
||||
"title": "DeepSeek's WEIRD Behavior: The cultural alignment of Large Language Models and the effects of prompt language and cultural prompting",
|
||||
"authors": "Authors: James Luther , Donald Brown",
|
||||
"dates": "Submitted 12 December, 2025; v1 submitted 10 December, 2025; originally announced December 2025."
|
||||
},
|
||||
{
|
||||
"id": "2512.02556",
|
||||
"title": "DeepSeek -V3.2: Pushing the Frontier of Open Large Language Models",
|
||||
"authors": "Authors: DeepSeek-AI , Aixin Liu , Aoxue Mei , Bangcai Lin , Bing Xue , Bingxuan Wang , Bingzheng Xu , Bochao Wu , Bowei Zhang , Chaofan Lin , Chen Dong , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chenhao Xu , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Erhang Li , Fangqi Zhou , Fangyun Lin , Fucong Dai , Guangbo Hao , et al. (239 additional authors not shown)",
|
||||
"dates": "Submitted 2 December, 2025; originally announced December 2025."
|
||||
},
|
||||
{
|
||||
"id": "2511.19333",
|
||||
"title": "Learning to Reason: Training LLMs with GPT-OSS or DeepSeek R1 Reasoning Traces",
|
||||
"authors": "Authors: Shaltiel Shmidman , Asher Fredman , Oleg Sudakov , Meriem Bendris",
|
||||
"dates": "Submitted 24 November, 2025; originally announced November 2025."
|
||||
},
|
||||
{
|
||||
"id": "2510.18234",
|
||||
"title": "DeepSeek -OCR: Contexts Optical Compression",
|
||||
"authors": "Authors: Haoran Wei , Yaofeng Sun , Yukun Li",
|
||||
"dates": "Submitted 20 October, 2025; originally announced October 2025."
|
||||
},
|
||||
{
|
||||
"id": "2510.15009",
|
||||
"title": "Can generative AI figure out figurative language? The influence of idioms on essay scoring by ChatGPT, Gemini, and Deepseek",
|
||||
"authors": "Authors: Enis Oğuz",
|
||||
"dates": "Submitted 16 October, 2025; originally announced October 2025."
|
||||
},
|
||||
{
|
||||
"id": "2509.23676",
|
||||
"title": "From Reasoning to Answer: Empirical, Attention-Based and Mechanistic Insights into Distilled DeepSeek R1 Models",
|
||||
"authors": "Authors: Jue Zhang , Qingwei Lin , Saravan Rajmohan , Dongmei Zhang",
|
||||
"dates": "Submitted 28 September, 2025; originally announced September 2025."
|
||||
},
|
||||
{
|
||||
"id": "2509.19346",
|
||||
"title": "Benchmarking ChatGPT and DeepSeek in April 2025: A Novel Dual Perspective Sentiment Analysis Using Lexicon-Based and Deep Learning Approaches",
|
||||
"authors": "Authors: Maryam Mahdi Alhusseini , Mohammad-Reza Feizi-Derakhshi",
|
||||
"dates": "Submitted 16 September, 2025; originally announced September 2025."
|
||||
},
|
||||
{
|
||||
"id": "2509.02036",
|
||||
"title": "DeepSeek performs better than other Large Language Models in Dental Cases",
|
||||
"authors": "Authors: Hexian Zhang , Xinyu Yan , Yanqi Yang , Lijian Jin , Ping Yang , Junwen Wang",
|
||||
"dates": "Submitted 2 September, 2025; originally announced September 2025."
|
||||
},
|
||||
{
|
||||
"id": "2508.21377",
|
||||
"title": "Challenges and Applications of Large Language Models: A Comparison of GPT and DeepSeek family of models",
|
||||
"authors": "Authors: Shubham Sharma , Sneha Tuli , Narendra Badam",
|
||||
"dates": "Submitted 29 August, 2025; originally announced August 2025."
|
||||
},
|
||||
{
|
||||
"id": "2508.10173",
|
||||
"title": "Benchmark-Driven Selection of AI: Evidence from DeepSeek -R1",
|
||||
"authors": "Authors: Petr Spelda , Vit Stritecky",
|
||||
"dates": "Submitted 13 August, 2025; originally announced August 2025."
|
||||
},
|
||||
{
|
||||
"id": "2507.19493",
|
||||
"title": "From Bench to Bedside: A DeepSeek -Powered AI System for Automated Chest Radiograph Interpretation in Clinical Practice",
|
||||
"authors": "Authors: Yaowei Bai , Ruiheng Zhang , Yu Lei , Jingfeng Yao , Shuguang Ju , Chaoyang Wang , Wei Yao , Yiwan Guo , Guilin Zhang , Chao Wan , Qian Yuan , Xuhua Duan , Xinggang Wang , Tao Sun , Yongchao Xu , Chuansheng Zheng , Huangxuan Zhao , Bo Du",
|
||||
"dates": "Submitted 31 May, 2025; originally announced July 2025."
|
||||
},
|
||||
{
|
||||
"id": "2507.17944",
|
||||
"title": "Evaluating the Performance of AI Text Detectors, Few-Shot and Chain-of-Thought Prompting Using DeepSeek Generated Text",
|
||||
"authors": "Authors: Hulayyil Alshammari , Praveen Rao",
|
||||
"dates": "Submitted 23 July, 2025; originally announced July 2025."
|
||||
},
|
||||
{
|
||||
"id": "2507.09955",
|
||||
"title": "DeepSeek : Paradigm Shifts and Technical Evolution in Large AI Models",
|
||||
"authors": "Authors: Luolin Xiong , Haofen Wang , Xi Chen , Lu Sheng , Yun Xiong , Jingping Liu , Yanghua Xiao , Huajun Chen , Qing-Long Han , Yang Tang",
|
||||
"dates": "Submitted 14 July, 2025; originally announced July 2025."
|
||||
},
|
||||
{
|
||||
"id": "2507.08621",
|
||||
"title": "A comprehensive study of LLM-based argument classification: from LLAMA through GPT-4o to Deepseek -R1",
|
||||
"authors": "Authors: Marcin Pietroń , Rafał Olszowski , Jakub Gomułka , Filip Gampel , Andrzej Tomski",
|
||||
"dates": "Submitted 24 July, 2025; v1 submitted 11 July, 2025; originally announced July 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.23128",
|
||||
"title": "Are Large Language Models Capable of Deep Relational Reasoning? Insights from DeepSeek -R1 and Benchmark Comparisons",
|
||||
"authors": "Authors: Chi Chiu So , Yueyue Sun , Jun-Min Wang , Siu Pang Yung , Anthony Wai Keung Loh , Chun Pong Chau",
|
||||
"dates": "Submitted 29 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.21562",
|
||||
"title": "FloorPlan- DeepSeek (FPDS): A multimodal approach to floorplan generation using vector-based next room prediction",
|
||||
"authors": "Authors: Jun Yin , Pengyu Zeng , Jing Zhong , Peilin Li , Miao Zhang , Ran Luo , Shuai Lu",
|
||||
"dates": "Submitted 2 August, 2025; v1 submitted 12 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.18543",
|
||||
"title": "SoK: A Comprehensive Security Analysis of Jailbreak Resilience in GPT and DeepSeek Models",
|
||||
"authors": "Authors: Xiaodong Wu , Xiangman Li , Qi Li , Lingshuang Liu , Jianbing Ni",
|
||||
"dates": "Submitted 25 May, 2026; v1 submitted 23 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.18501",
|
||||
"title": "Comparative Evaluation of ChatGPT and DeepSeek Across Key NLP Tasks: Strengths, Weaknesses, and Domain-Specific Performance",
|
||||
"authors": "Authors: Wael Etaiwi , Bushra Alhijawi",
|
||||
"dates": "Submitted 5 July, 2025; v1 submitted 23 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.14104",
|
||||
"title": "Innovating China's Intangible Cultural Heritage with DeepSeek + MidJourney: The Case of Yangliuqing theme Woodblock Prints",
|
||||
"authors": "Authors: RuiKun Yang , ZhongLiang Wei , Longdi Xian",
|
||||
"dates": "Submitted 16 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.12349",
|
||||
"title": "Information Suppression in Large Language Models: Auditing, Quantifying, and Characterizing Censorship in DeepSeek",
|
||||
"authors": "Authors: Peiran Qiu , Siyi Zhou , Emilio Ferrara",
|
||||
"dates": "Submitted 14 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.05921",
|
||||
"title": "Multi-Modal Large Models Based Beam Prediction: An Example Empowered by DeepSeek",
|
||||
"authors": "Authors: Yizhu Zhao , Li Yu , Lianzheng Shi , Jianhua Zhang , Guangyi Liu",
|
||||
"dates": "Submitted 6 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.02523",
|
||||
"title": "Hardware-Centric Analysis of DeepSeek's Multi-Head Latent Attention",
|
||||
"authors": "Authors: Robin Geens , Marian Verhelst",
|
||||
"dates": "Submitted 3 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.01814",
|
||||
"title": "Analysis of LLM Bias (Chinese Propaganda & Anti-US Sentiment) in DeepSeek -R1 vs. ChatGPT o3-mini-high",
|
||||
"authors": "Authors: PeiHsuan Huang , ZihWei Lin , Simon Imbot , WenCheng Fu , Ethan Tu",
|
||||
"dates": "Submitted 2 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.01257",
|
||||
"title": "DeepSeek in Healthcare: A Survey of Capabilities, Risks, and Clinical Applications of Open-Source Large Language Models",
|
||||
"authors": "Authors: Jiancheng Ye , Sophie Bronstein , Jiarui Hai , Malak Abu Hashish",
|
||||
"dates": "Submitted 1 June, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2506.00312",
|
||||
"title": "An evaluation of LLMs for generating movie reviews: GPT-4o, Gemini-2.0 and DeepSeek -V3",
|
||||
"authors": "Authors: Brendan Sands , Yining Wang , Chenhao Xu , Yuxuan Zhou , Lai Wei , Rohitash Chandra",
|
||||
"dates": "Submitted 30 May, 2025; originally announced June 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.18059",
|
||||
"title": "Assessing the performance of 8 AI chatbots in bibliographic reference retrieval: Grok and DeepSeek outperform ChatGPT, but none are fully accurate",
|
||||
"authors": "Authors: Álvaro Cabezas-Clavijo , Pavel Sidorenko-Bautista",
|
||||
"dates": "Submitted 23 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.17702",
|
||||
"title": "Seek-CAD: A Self-refined Generative Modeling for 3D Parametric CAD Using Local Inference via DeepSeek",
|
||||
"authors": "Authors: Xueyang Li , Jiahao Li , Yu Song , Yunzhong Lou , Xiangdong Zhou",
|
||||
"dates": "Submitted 28 February, 2026; v1 submitted 23 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.17112",
|
||||
"title": "Cultural Value Alignment in Large Language Models: A Prompt-based Analysis of Schwartz Values in Gemini, ChatGPT, and DeepSeek",
|
||||
"authors": "Authors: Robin Segerer",
|
||||
"dates": "Submitted 21 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.12625",
|
||||
"title": "R1dacted: Investigating Local Censorship in DeepSeek's R1 Language Model",
|
||||
"authors": "Authors: Ali Naseh , Harsh Chaudhari , Jaechul Roh , Mingshi Wu , Alina Oprea , Amir Houmansadr",
|
||||
"dates": "Submitted 18 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.09343",
|
||||
"title": "Insights into DeepSeek -V3: Scaling Challenges and Reflections on Hardware for AI Architectures",
|
||||
"authors": "Authors: Chenggang Zhao , Chengqi Deng , Chong Ruan , Damai Dai , Huazuo Gao , Jiashi Li , Liyue Zhang , Panpan Huang , Shangyan Zhou , Shirong Ma , Wenfeng Liang , Ying He , Yuqing Wang , Yuxuan Liu , Y. X. Wei",
|
||||
"dates": "Submitted 22 December, 2025; v1 submitted 14 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.02390",
|
||||
"title": "Quantitative Analysis of Performance Drop in DeepSeek Model Quantization",
|
||||
"authors": "Authors: Enbo Zhao , Yi Shen , Shuming Shi , Jieyun Huang , Zhihao Chen , Ning Wang , Siqi Xiao , Jian Zhang , Kai Wang , Shiguo Lian",
|
||||
"dates": "Submitted 13 June, 2025; v1 submitted 5 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.00551",
|
||||
"title": "100 Days After DeepSeek -R1: A Survey on Replication Studies and More Directions for Reasoning Language Models",
|
||||
"authors": "Authors: Chong Zhang , Yue Deng , Xiang Lin , Bin Wang , Dianwen Ng , Hai Ye , Xingxuan Li , Yao Xiao , Zhanfeng Mo , Qi Zhang , Lidong Bing",
|
||||
"dates": "Submitted 15 May, 2025; v1 submitted 1 May, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2505.00025",
|
||||
"title": "A Method for the Architecture of a Medical Vertical Large Language Model Based on Deepseek R1",
|
||||
"authors": "Authors: Mingda Zhang , Jianglong Qin",
|
||||
"dates": "Submitted 22 July, 2025; v1 submitted 25 April, 2025; originally announced May 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.21801",
|
||||
"title": "DeepSeek -Prover-V2: Advancing Formal Mathematical Reasoning via Reinforcement Learning for Subgoal Decomposition",
|
||||
"authors": "Authors: Z. Z. Ren , Zhihong Shao , Junxiao Song , Huajian Xin , Haocheng Wang , Wanjia Zhao , Liyue Zhang , Zhe Fu , Qihao Zhu , Dejian Yang , Z. F. Wu , Zhibin Gou , Shirong Ma , Hongxuan Tang , Yuxuan Liu , Wenjun Gao , Daya Guo , Chong Ruan",
|
||||
"dates": "Submitted 18 July, 2025; v1 submitted 30 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.16027",
|
||||
"title": "Benchmarking LLM for Code Smells Detection: OpenAI GPT-4.0 vs DeepSeek -V3",
|
||||
"authors": "Authors: Ahmed R. Sadik , Siddhata Govind",
|
||||
"dates": "Submitted 22 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.10081",
|
||||
"title": "RealSafe-R1: Safety-Aligned DeepSeek -R1 without Compromising Reasoning Capability",
|
||||
"authors": "Authors: Yichi Zhang , Zihao Zeng , Dongbai Li , Yao Huang , Zhijie Deng , Yinpeng Dong",
|
||||
"dates": "Submitted 14 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.08120",
|
||||
"title": "DeepSeek -R1 vs. o3-mini: How Well can Reasoning LLMs Evaluate MT and Summarization?",
|
||||
"authors": "Authors: Daniil Larionov , Sotaro Takeshita , Ran Zhang , Yanran Chen , Christoph Leiter , Zhipin Wang , Christian Greisinger , Steffen Eger",
|
||||
"dates": "Submitted 30 May, 2025; v1 submitted 10 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.07128",
|
||||
"title": "DeepSeek -R1 Thoughtology: Let's think about LLM Reasoning",
|
||||
"authors": "Authors: Sara Vera Marjanović , Arkil Patel , Vaibhav Adlakha , Milad Aghajohari , Parishad BehnamGhader , Mehar Bhatia , Aditi Khandelwal , Austin Kraft , Benno Krojer , Xing Han Lù , Nicholas Meade , Dongchan Shin , Amirhossein Kazemnejad , Gaurav Kamath , Marius Mosbach , Karolina Stańczak , Siva Reddy",
|
||||
"dates": "Submitted 15 January, 2026; v1 submitted 1 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.03665",
|
||||
"title": "LLM & HPC:Benchmarking DeepSeek's Performance in High-Performance Computing Tasks",
|
||||
"authors": "Authors: Noujoud Nader , Patrick Diehl , Steve Brandt , Hartmut Kaiser",
|
||||
"dates": "Submitted 28 April, 2025; v1 submitted 15 March, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.03219",
|
||||
"title": "From ChatGPT to DeepSeek AI: A Comprehensive Analysis of Evolution, Deviation, and Future Implications in AI-Language Models",
|
||||
"authors": "Authors: Simrandeep Singh , Shreya Bansal , Abdulmotaleb El Saddik , Mukesh Saini",
|
||||
"dates": "Submitted 4 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.02888",
|
||||
"title": "A Status Quo Investigation of Large Language Models towards Cost-Effective CFD Automation with OpenFOAMGPT: ChatGPT vs. Qwen vs. Deepseek",
|
||||
"authors": "Authors: Wenkang Wang , Ran Xu , Jingsen Feng , Qingfu Zhang , Xu Chu",
|
||||
"dates": "Submitted 2 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.00421",
|
||||
"title": "Can LLMs Assist Computer Education? an Empirical Case Study of DeepSeek",
|
||||
"authors": "Authors: Dongfu Xiao , Chen Gao , Zhengquan Luo , Chi Liu , Sheng Shen",
|
||||
"dates": "Submitted 1 April, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2504.00016",
|
||||
"title": "Medical Reasoning in LLMs: An In-Depth Analysis of DeepSeek R1",
|
||||
"authors": "Authors: Birger Moell , Fredrik Sand Aronsson , Sanian Akbar",
|
||||
"dates": "Submitted 27 March, 2025; originally announced April 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.23130",
|
||||
"title": "Can DeepSeek Reason Like a Surgeon? An Empirical Evaluation for Vision-Language Understanding in Robotic-Assisted Surgery",
|
||||
"authors": "Authors: Boyi Ma , Yanguang Zhao , Jie Wang , Guankun Wang , Kun Yuan , Tong Chen , Long Bai , Hongliang Ren",
|
||||
"dates": "Submitted 3 April, 2025; v1 submitted 29 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.17365",
|
||||
"title": "How Effective Is Constitutional AI in Small LLMs? A Study on DeepSeek -R1 and Its Peers",
|
||||
"authors": "Authors: Antonio-Gabriel Chacón Menke , Phan Xuan Tan",
|
||||
"dates": "Submitted 11 April, 2025; v1 submitted 1 February, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.16529",
|
||||
"title": "Safety Evaluation and Enhancement of DeepSeek Models in Chinese Contexts",
|
||||
"authors": "Authors: Wenjing Zhang , Xuejiao Lei , Zhaoxiang Liu , Limin Han , Jiaojiao Zhao , Junting Guo , Zhenhong Long , Shu Yang , Meijuan An , Beibei Huang , Rongjia Du , Ning Wang , Kai Wang , Shiguo Lian",
|
||||
"dates": "Submitted 16 May, 2025; v1 submitted 18 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.16304",
|
||||
"title": "Bridging Technology and Humanities: Evaluating the Impact of Large Language Models on Social Sciences Research with DeepSeek -R1",
|
||||
"authors": "Authors: Peiran Gu , Fuhao Duan , Wenhao Li , Bochen Xu , Ying Cai , Teng Yao , Chenxun Zhuo , Tianming Liu , Bao Ge",
|
||||
"dates": "Submitted 15 April, 2025; v1 submitted 20 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.16040",
|
||||
"title": "Evaluating Test-Time Scaling LLMs for Legal Reasoning: OpenAI o1, DeepSeek -R1, and Beyond",
|
||||
"authors": "Authors: Yinghao Hu , Yaoyao Yu , Leilei Gan , Bin Wei , Kun Kuang , Fei Wu",
|
||||
"dates": "Submitted 10 November, 2025; v1 submitted 20 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.15092",
|
||||
"title": "Towards Understanding the Safety Boundaries of DeepSeek Models: Evaluation and Findings",
|
||||
"authors": "Authors: Zonghao Ying , Guangyi Zheng , Yongxin Huang , Deyue Zhang , Wenxin Zhang , Quanchen Zou , Aishan Liu , Xianglong Liu , Dacheng Tao",
|
||||
"dates": "Submitted 19 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.13549",
|
||||
"title": "A Showdown of ChatGPT vs DeepSeek in Solving Programming Tasks",
|
||||
"authors": "Authors: Ronas Shakya , Sam Urmian , Mohammad Khalil",
|
||||
"dates": "Submitted 16 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.11655",
|
||||
"title": "Explainable Sentiment Analysis with DeepSeek -R1: Performance, Efficiency, and Few-Shot Learning",
|
||||
"authors": "Authors: Donghao Huang , Zhaoxia Wang",
|
||||
"dates": "Submitted 3 February, 2026; v1 submitted 3 February, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.11486",
|
||||
"title": "A Review of DeepSeek Models' Key Innovative Techniques",
|
||||
"authors": "Authors: Chengen Wang , Murat Kantarcioglu",
|
||||
"dates": "Submitted 14 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.11068",
|
||||
"title": "DeepSeek Powered Solid Dosage Formulation Design and Development",
|
||||
"authors": "Authors: Leqi Lin , Xingyu Zhou , Kaiyuan Yang , Xizhong Chen",
|
||||
"dates": "Submitted 21 March, 2025; v1 submitted 14 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.10486",
|
||||
"title": "LLMs in Disease Diagnosis: A Comparative Study of DeepSeek -R1 and O3 Mini Across Chronic Health Conditions",
|
||||
"authors": "Authors: Gaurav Kumar Gupta , Pranal Pande , Nirajan Acharya , Aniket Kumar Singh , Suman Niroula",
|
||||
"dates": "Submitted 19 June, 2025; v1 submitted 13 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.09956",
|
||||
"title": "DeepSeek -Inspired Exploration of RL-based LLMs and Synergy with Wireless Networks: A Survey",
|
||||
"authors": "Authors: Yu Qiao , Phuong-Nam Tran , Ji Su Yoon , Loc X. Nguyen , Eui-Nam Huh , Dusit Niyato , Choong Seon Hong",
|
||||
"dates": "Submitted 20 October, 2025; v1 submitted 12 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.04783",
|
||||
"title": "Comparative Analysis Based on DeepSeek , ChatGPT, and Google Gemini: Features, Techniques, Performance, Future Prospects",
|
||||
"authors": "Authors: Anichur Rahman , Shahariar Hossain Mahir , Md Tanjum An Tashrif , Airin Afroj Aishi , Md Ahsan Karim , Dipanjali Kundu , Tanoy Debnath , Md. Abul Ala Moududi , MD. Zunead Abedin Eidmum",
|
||||
"dates": "Submitted 25 February, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.04765",
|
||||
"title": "Generative AI in Academic Writing: A Comparison of DeepSeek , Qwen, ChatGPT, Gemini, Llama, Mistral, and Gemma",
|
||||
"authors": "Authors: Omer Aydin , Enis Karaarslan , Fatih Safa Erenay , Nebojsa Bacanin",
|
||||
"dates": "Submitted 1 May, 2025; v1 submitted 11 February, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.02032",
|
||||
"title": "Comparative Analysis of OpenAI GPT-4o and DeepSeek R1 for Scientific Text Categorization Using Prompt Engineering",
|
||||
"authors": "Authors: Aniruddha Maiti , Samuel Adewumi , Temesgen Alemayehu Tikure , Zichun Wang , Niladri Sengupta , Anastasiia Sukhanova , Ananya Jana",
|
||||
"dates": "Submitted 3 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.01923",
|
||||
"title": "Output Length Effect on DeepSeek -R1's Safety in Forced Thinking",
|
||||
"authors": "Authors: Xuying Li , Zhuo Li , Yuji Kosuga , Victor Bian",
|
||||
"dates": "Submitted 2 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2503.00624",
|
||||
"title": "An evaluation of DeepSeek Models in Biomedical Natural Language Processing",
|
||||
"authors": "Authors: Zaifu Zhan , Shuang Zhou , Huixue Zhou , Jiawen Deng , Yu Hou , Jeremy Yeung , Rui Zhang",
|
||||
"dates": "Submitted 1 March, 2025; originally announced March 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.18513",
|
||||
"title": "Analyzing User Perceptions of Large Language Models (LLMs) on Reddit: Sentiment and Topic Modeling of ChatGPT and DeepSeek Discussions",
|
||||
"authors": "Authors: Krishnaveni Katta",
|
||||
"dates": "Submitted 22 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.18505",
|
||||
"title": "Comprehensive Analysis of Transparency and Accessibility of ChatGPT, DeepSeek , And other SoTA Large Language Models",
|
||||
"authors": "Authors: Ranjan Sapkota , Shaina Raza , Manoj Karkee",
|
||||
"dates": "Submitted 21 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.18467",
|
||||
"title": "ChatGPT vs. DeepSeek : A Comparative Study on AI-Based Code Generation",
|
||||
"authors": "Authors: Md Motaleb Hossen Manik",
|
||||
"dates": "Submitted 30 January, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.18210",
|
||||
"title": "From ChatGPT to DeepSeek : Can LLMs Simulate Humanity?",
|
||||
"authors": "Authors: Qian Wang , Zhenheng Tang , Bingsheng He",
|
||||
"dates": "Submitted 25 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.17947",
|
||||
"title": "DeepSeek -R1 Outperforms Gemini 2.0 Pro, OpenAI o1, and o3-mini in Bilingual Complex Ophthalmology Reasoning",
|
||||
"authors": "Authors: Pusheng Xu , Yue Wu , Kai Jin , Xiaolan Chen , Mingguang He , Danli Shi",
|
||||
"dates": "Submitted 25 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.17764",
|
||||
"title": "DeepSeek vs. ChatGPT vs. Claude: A Comparative Study for Scientific Computing and Scientific Machine Learning Tasks",
|
||||
"authors": "Authors: Qile Jiang , Zhiwei Gao , George Em Karniadakis",
|
||||
"dates": "Submitted 8 March, 2025; v1 submitted 24 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.17487",
|
||||
"title": "User Intent to Use DeepSeek for Healthcare Purposes and their Trust in the Large Language Model: Multinational Survey Study",
|
||||
"authors": "Authors: Avishek Choudhury , Yeganeh Shahsavar , Hamid Shamszare",
|
||||
"dates": "Submitted 2 March, 2025; v1 submitted 18 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.16732",
|
||||
"title": "DeepSeek reshaping healthcare in China's tertiary hospitals",
|
||||
"authors": "Authors: Jishizhan Chen , Qingzeng Zhang",
|
||||
"dates": "Submitted 27 February, 2025; v1 submitted 23 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.16428",
|
||||
"title": "Visual Reasoning Evaluation of Grok, Deepseek Janus, Gemini, Qwen, Mistral, and ChatGPT",
|
||||
"authors": "Authors: Nidhal Jegham , Marwan Abdelatti , Abdeltawab Hendawi",
|
||||
"dates": "Submitted 22 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.14926",
|
||||
"title": "DeepSeek -V3, GPT-4, Phi-4, and LLaMA-3.3 generate correct code for LoRaWAN-related engineering tasks",
|
||||
"authors": "Authors: Daniel Fernandes , João P. Matos-Carvalho , Carlos M. Fernandes , Nuno Fachada",
|
||||
"dates": "Submitted 7 April, 2025; v1 submitted 19 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.14837",
|
||||
"title": "Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs",
|
||||
"authors": "Authors: Tao Ji , Bin Guo , Yuanbin Wu , Qipeng Guo , Lixing Shen , Zhan Chen , Xipeng Qiu , Qi Zhang , Tao Gui",
|
||||
"dates": "Submitted 3 October, 2025; v1 submitted 20 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.12893",
|
||||
"title": "H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek -R1, and Gemini 2.0 Flash Thinking",
|
||||
"authors": "Authors: Martin Kuo , Jianyi Zhang , Aolin Ding , Qinsi Wang , Louis DiValentin , Yujia Bao , Wei Wei , Hai Li , Yiran Chen",
|
||||
"dates": "Submitted 26 February, 2025; v1 submitted 18 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.12825",
|
||||
"title": "Reasoning and the Trusting Behavior of DeepSeek and GPT: An Experiment Revealing Hidden Fault Lines in Large Language Models",
|
||||
"authors": "Authors: Rubing Li , João Sedoc , Arun Sundararajan",
|
||||
"dates": "Submitted 19 February, 2025; v1 submitted 18 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.11347",
|
||||
"title": "Evaluating the Performance of the DeepSeek Model in Confidential Computing Environment",
|
||||
"authors": "Authors: Ben Dong , Qian Wang",
|
||||
"dates": "Submitted 16 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.11164",
|
||||
"title": "Quantifying the Capability Boundary of DeepSeek Models: An Application-Driven Performance Analysis",
|
||||
"authors": "Authors: Kaikai Zhao , Zhaoxiang Liu , Xuejiao Lei , Jiaojiao Zhao , Zhenhong Long , Zipeng Wang , Ning Wang , Meijuan An , Qingliang Meng , Peijun Yang , Minjie Hua , Chaoyang Ma , Wen Liu , Kai Wang , Shiguo Lian",
|
||||
"dates": "Submitted 15 May, 2025; v1 submitted 16 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.11137",
|
||||
"title": "Safety Evaluation of DeepSeek Models in Chinese Contexts",
|
||||
"authors": "Authors: Wenjing Zhang , Xuejiao Lei , Zhaoxiang Liu , Ning Wang , Zhenhong Long , Peijun Yang , Jiaojiao Zhao , Minjie Hua , Chaoyang Ma , Kai Wang , Shiguo Lian",
|
||||
"dates": "Submitted 7 May, 2025; v1 submitted 16 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.11108",
|
||||
"title": "Knowledge Graph-Driven Retrieval-Augmented Generation: Integrating Deepseek -R1 with Weaviate for Advanced Chatbot Applications",
|
||||
"authors": "Authors: Alexandru Lecu , Adrian Groza , Lezan Hawizy",
|
||||
"dates": "Submitted 16 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.11096",
|
||||
"title": "Mixture of Tunable Experts -- Behavior Modification of DeepSeek -R1 at Inference Time",
|
||||
"authors": "Authors: Robert Dahlke , Henrik Klagges , Dan Zecha , Benjamin Merkel , Sven Rohr , Fabian Klemm",
|
||||
"dates": "Submitted 16 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.10299",
|
||||
"title": "Open-Source AI-Powered Optimization in Scalene: Advancing Python Performance Profiling with DeepSeek -R1 and LLaMA 3.2",
|
||||
"authors": "Authors: Saem Hasan , Sanju Basak",
|
||||
"dates": "Submitted 14 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.10008",
|
||||
"title": "ChatGPT and Deepseek : Can They Predict the Stock Market and Macroeconomy?",
|
||||
"authors": "Authors: Jian Chen , Guohao Tang , Guofu Zhou , Wu Zhu",
|
||||
"dates": "Submitted 14 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.07905",
|
||||
"title": "DeepSeek on a Trip: Inducing Targeted Visual Hallucinations via Representation Vulnerabilities",
|
||||
"authors": "Authors: Chashi Mahiul Islam , Samuel Jacob Chacko , Preston Horne , Xiuwen Liu",
|
||||
"dates": "Submitted 11 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.07846",
|
||||
"title": "Memory Analysis on the Training Course of DeepSeek Models",
|
||||
"authors": "Authors: Ping Zhang , Lei Su",
|
||||
"dates": "Submitted 11 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.07393",
|
||||
"title": "FinRL- DeepSeek : LLM-Infused Risk-Sensitive Reinforcement Learning for Trading Agents",
|
||||
"authors": "Authors: Mostapha Benhenda",
|
||||
"dates": "Submitted 11 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.05769",
|
||||
"title": "Digital Twin Buildings: 3D Modeling, GIS Integration, and Visual Descriptions Using Gaussian Splatting, ChatGPT/ Deepseek , and Google Maps Platform",
|
||||
"authors": "Authors: Kyle Gao , Dening Lu , Liangzhi Li , Nan Chen , Hongjie He , Linlin Xu , Jonathan Li",
|
||||
"dates": "Submitted 20 April, 2025; v1 submitted 8 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.03688",
|
||||
"title": "A Comparison of DeepSeek and Other LLMs",
|
||||
"authors": "Authors: Tianchen Gao , Jiashun Jin , Zheng Tracy Ke , Gabriel Moryoussef",
|
||||
"dates": "Submitted 25 December, 2025; v1 submitted 5 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.02523",
|
||||
"title": "Brief analysis of DeepSeek R1 and its implications for Generative AI",
|
||||
"authors": "Authors: Sarah Mercer , Samuel Spillard , Daniel P. Martin",
|
||||
"dates": "Submitted 7 February, 2025; v1 submitted 4 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2502.01225",
|
||||
"title": "The dark deep side of DeepSeek : Fine-tuning attacks against the safety alignment of CoT-enabled models",
|
||||
"authors": "Authors: Zhiyuan Xu , Joseph Gardiner , Sana Belguith",
|
||||
"dates": "Submitted 3 February, 2025; originally announced February 2025."
|
||||
},
|
||||
{
|
||||
"id": "2501.18576",
|
||||
"title": "Token-Hungry, Yet Precise: DeepSeek R1 Highlights the Need for Multi-Step Reasoning Over Speed in MATH",
|
||||
"authors": "Authors: Evgenii Evstafev",
|
||||
"dates": "Submitted 30 January, 2025; originally announced January 2025."
|
||||
},
|
||||
{
|
||||
"id": "2501.18438",
|
||||
"title": "o3-mini vs DeepSeek -R1: Which One is Safer?",
|
||||
"authors": "Authors: Aitor Arrieta , Miriam Ugarte , Pablo Valle , José Antonio Parejo , Sergio Segura",
|
||||
"dates": "Submitted 31 January, 2025; v1 submitted 30 January, 2025; originally announced January 2025."
|
||||
},
|
||||
{
|
||||
"id": "2501.17030",
|
||||
"title": "Challenges in Ensuring AI Safety in DeepSeek -R1 Models: The Shortcomings of Reinforcement Learning Strategies",
|
||||
"authors": "Authors: Manojkumar Parmar , Yuvaraj Govindarajulu",
|
||||
"dates": "Submitted 28 January, 2025; originally announced January 2025."
|
||||
},
|
||||
{
|
||||
"id": "2501.12948",
|
||||
"title": "DeepSeek -R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning",
|
||||
"authors": "Authors: DeepSeek-AI , Daya Guo , Dejian Yang , Haowei Zhang , Junxiao Song , Peiyi Wang , Qihao Zhu , Runxin Xu , Ruoyu Zhang , Shirong Ma , Xiao Bi , Xiaokang Zhang , Xingkai Yu , Yu Wu , Z. F. Wu , Zhibin Gou , Zhihong Shao , Zhuoshu Li , Ziyi Gao , Aixin Liu , Bing Xue , Bingxuan Wang , Bochao Wu , Bei Feng , Chengda Lu , et al. (175 additional authors not shown)",
|
||||
"dates": "Submitted 3 January, 2026; v1 submitted 22 January, 2025; originally announced January 2025."
|
||||
},
|
||||
{
|
||||
"id": "2501.08156",
|
||||
"title": "Are DeepSeek R1 And Other Reasoning Models More Faithful?",
|
||||
"authors": "Authors: James Chua , Owain Evans",
|
||||
"dates": "Submitted 15 July, 2025; v1 submitted 14 January, 2025; originally announced January 2025."
|
||||
},
|
||||
{
|
||||
"id": "2412.19437",
|
||||
"title": "DeepSeek -V3 Technical Report",
|
||||
"authors": "Authors: DeepSeek-AI , Aixin Liu , Bei Feng , Bing Xue , Bingxuan Wang , Bochao Wu , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chenyu Zhang , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Dongjie Ji , Erhang Li , Fangyun Lin , Fucong Dai , Fuli Luo , Guangbo Hao , Guanting Chen , Guowei Li , H. Zhang , Han Bao , et al. (175 additional authors not shown)",
|
||||
"dates": "Submitted 18 February, 2025; v1 submitted 26 December, 2024; originally announced December 2024."
|
||||
},
|
||||
{
|
||||
"id": "2412.10302",
|
||||
"title": "DeepSeek -VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding",
|
||||
"authors": "Authors: Zhiyu Wu , Xiaokang Chen , Zizheng Pan , Xingchao Liu , Wen Liu , Damai Dai , Huazuo Gao , Yiyang Ma , Chengyue Wu , Bingxuan Wang , Zhenda Xie , Yu Wu , Kai Hu , Jiawei Wang , Yaofeng Sun , Yukun Li , Yishi Piao , Kang Guan , Aixin Liu , Xin Xie , Yuxiang You , Kai Dong , Xingkai Yu , Haowei Zhang , Liang Zhao , et al. (2 additional authors not shown)",
|
||||
"dates": "Submitted 13 December, 2024; originally announced December 2024."
|
||||
},
|
||||
{
|
||||
"id": "2410.01306",
|
||||
"title": "Emotion-Aware Embedding Fusion in LLMs (Flan-T5, LLAMA 2, DeepSeek -R1, and ChatGPT 4) for Intelligent Response Generation",
|
||||
"authors": "Authors: Abdur Rasool , Muhammad Irfan Shahzad , Hafsa Aslam , Vincent Chan , Muhammad Ali Arshad",
|
||||
"dates": "Submitted 11 March, 2025; v1 submitted 2 October, 2024; originally announced October 2024."
|
||||
},
|
||||
{
|
||||
"id": "2408.08152",
|
||||
"title": "DeepSeek -Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search",
|
||||
"authors": "Authors: Huajian Xin , Z. Z. Ren , Junxiao Song , Zhihong Shao , Wanjia Zhao , Haocheng Wang , Bo Liu , Liyue Zhang , Xuan Lu , Qiushi Du , Wenjun Gao , Qihao Zhu , Dejian Yang , Zhibin Gou , Z. F. Wu , Fuli Luo , Chong Ruan",
|
||||
"dates": "Submitted 15 August, 2024; originally announced August 2024."
|
||||
},
|
||||
{
|
||||
"id": "2406.11931",
|
||||
"title": "DeepSeek -Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence",
|
||||
"authors": "Authors: DeepSeek-AI , Qihao Zhu , Daya Guo , Zhihong Shao , Dejian Yang , Peiyi Wang , Runxin Xu , Y. Wu , Yukun Li , Huazuo Gao , Shirong Ma , Wangding Zeng , Xiao Bi , Zihui Gu , Hanwei Xu , Damai Dai , Kai Dong , Liyue Zhang , Yishi Piao , Zhibin Gou , Zhenda Xie , Zhewen Hao , Bingxuan Wang , Junxiao Song , Deli Chen , et al. (15 additional authors not shown)",
|
||||
"dates": "Submitted 17 June, 2024; originally announced June 2024."
|
||||
},
|
||||
{
|
||||
"id": "2405.14333",
|
||||
"title": "DeepSeek -Prover: Advancing Theorem Proving in LLMs through Large-Scale Synthetic Data",
|
||||
"authors": "Authors: Huajian Xin , Daya Guo , Zhihong Shao , Zhizhou Ren , Qihao Zhu , Bo Liu , Chong Ruan , Wenda Li , Xiaodan Liang",
|
||||
"dates": "Submitted 23 May, 2024; originally announced May 2024."
|
||||
},
|
||||
{
|
||||
"id": "2405.04434",
|
||||
"title": "DeepSeek -V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model",
|
||||
"authors": "Authors: DeepSeek-AI , Aixin Liu , Bei Feng , Bin Wang , Bingxuan Wang , Bo Liu , Chenggang Zhao , Chengqi Dengr , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Dongjie Ji , Erhang Li , Fangyun Lin , Fuli Luo , Guangbo Hao , Guanting Chen , Guowei Li , H. Zhang , Hanwei Xu , Hao Yang , Haowei Zhang , Honghui Ding , et al. (132 additional authors not shown)",
|
||||
"dates": "Submitted 19 June, 2024; v1 submitted 7 May, 2024; originally announced May 2024."
|
||||
},
|
||||
{
|
||||
"id": "2403.05525",
|
||||
"title": "DeepSeek -VL: Towards Real-World Vision-Language Understanding",
|
||||
"authors": "Authors: Haoyu Lu , Wen Liu , Bo Zhang , Bingxuan Wang , Kai Dong , Bo Liu , Jingxiang Sun , Tongzheng Ren , Zhuoshu Li , Hao Yang , Yaofeng Sun , Chengqi Deng , Hanwei Xu , Zhenda Xie , Chong Ruan",
|
||||
"dates": "Submitted 11 March, 2024; v1 submitted 8 March, 2024; originally announced March 2024."
|
||||
},
|
||||
{
|
||||
"id": "2401.14196",
|
||||
"title": "DeepSeek -Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence",
|
||||
"authors": "Authors: Daya Guo , Qihao Zhu , Dejian Yang , Zhenda Xie , Kai Dong , Wentao Zhang , Guanting Chen , Xiao Bi , Y. Wu , Y. K. Li , Fuli Luo , Yingfei Xiong , Wenfeng Liang",
|
||||
"dates": "Submitted 26 January, 2024; v1 submitted 25 January, 2024; originally announced January 2024."
|
||||
},
|
||||
{
|
||||
"id": "2401.02954",
|
||||
"title": "DeepSeek LLM: Scaling Open-Source Language Models with Longtermism",
|
||||
"authors": "Authors: DeepSeek-AI , : , Xiao Bi , Deli Chen , Guanting Chen , Shanhuang Chen , Damai Dai , Chengqi Deng , Honghui Ding , Kai Dong , Qiushi Du , Zhe Fu , Huazuo Gao , Kaige Gao , Wenjun Gao , Ruiqi Ge , Kang Guan , Daya Guo , Jianzhong Guo , Guangbo Hao , Zhewen Hao , Ying He , Wenjie Hu , Panpan Huang , Erhang Li , et al. (63 additional authors not shown)",
|
||||
"dates": "Submitted 5 January, 2024; originally announced January 2024."
|
||||
},
|
||||
{
|
||||
"id": "1801.03406",
|
||||
"title": "DeepSeek : Content Based Image Search & Retrieval",
|
||||
"authors": "Authors: Tanya Piplani , David Bamman",
|
||||
"dates": "Submitted 11 January, 2018; v1 submitted 9 January, 2018; originally announced January 2018."
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,128 @@
|
||||
2504.00421 Can LLMs Assist Computer Education? an Empirical Case Study of DeepSee Submitted 1 April, 2025; originally an
|
||||
2506.01257 DeepSeek in Healthcare: A Survey of Capabilities, Risks, and Clinical Submitted 1 June, 2025; originally ann
|
||||
2503.00624 An evaluation of DeepSeek Models in Biomedical Natural Language Proces Submitted 1 March, 2025; originally an
|
||||
2503.04765 Generative AI in Academic Writing: A Comparison of DeepSeek , Qwen, Ch Submitted 1 May, 2025; v1 submitted 11
|
||||
2503.16040 Evaluating Test-Time Scaling LLMs for Legal Reasoning: OpenAI o1, Deep Submitted 10 November, 2025; v1 submit
|
||||
2503.17365 How Effective Is Constitutional AI in Small LLMs? A Study on DeepSeek Submitted 11 April, 2025; v1 submitted
|
||||
2512.10576 ESS: An Offload-Centric Latent-Cache Management Architecture for DeepS Submitted 11 December, 2025; originall
|
||||
2502.07905 DeepSeek on a Trip: Inducing Targeted Visual Hallucinations via Repres Submitted 11 February, 2025; originall
|
||||
2502.07846 Memory Analysis on the Training Course of DeepSeek Models Submitted 11 February, 2025; originall
|
||||
2502.07393 FinRL- DeepSeek : LLM-Infused Risk-Sensitive Reinforcement Learning fo Submitted 11 February, 2025; originall
|
||||
1801.03406 DeepSeek : Content Based Image Search & Retrieval Submitted 11 January, 2018; v1 submitt
|
||||
2403.05525 DeepSeek -VL: Towards Real-World Vision-Language Understanding Submitted 11 March, 2024; v1 submitted
|
||||
2410.01306 Emotion-Aware Embedding Fusion in LLMs (Flan-T5, LLAMA 2, DeepSeek -R1 Submitted 11 March, 2025; v1 submitted
|
||||
2605.10234 Social Policy of Large Language Models: How GPT, Claude, DeepSeek and Submitted 11 May, 2026; originally ann
|
||||
2512.09772 DeepSeek's WEIRD Behavior: The cultural alignment of Large Language Mo Submitted 12 December, 2025; v1 submit
|
||||
2508.10173 Benchmark-Driven Selection of AI: Evidence from DeepSeek -R1 Submitted 13 August, 2025; originally
|
||||
2412.10302 DeepSeek -VL2: Mixture-of-Experts Vision-Language Models for Advanced Submitted 13 December, 2024; originall
|
||||
2505.02390 Quantitative Analysis of Performance Drop in DeepSeek Model Quantizati Submitted 13 June, 2025; v1 submitted
|
||||
2504.10081 RealSafe-R1: Safety-Aligned DeepSeek -R1 without Compromising Reasonin Submitted 14 April, 2025; originally a
|
||||
2604.12548 DeepSeek Robustness Against Semantic-Character Dual-Space Mutated Prom Submitted 14 April, 2026; originally a
|
||||
2502.10299 Open-Source AI-Powered Optimization in Scalene: Advancing Python Perfo Submitted 14 February, 2025; originall
|
||||
2502.10008 ChatGPT and Deepseek : Can They Predict the Stock Market and Macroecon Submitted 14 February, 2025; originall
|
||||
2507.09955 DeepSeek : Paradigm Shifts and Technical Evolution in Large AI Models Submitted 14 July, 2025; originally an
|
||||
2506.12349 Information Suppression in Large Language Models: Auditing, Quantifyin Submitted 14 June, 2025; originally an
|
||||
2503.11486 A Review of DeepSeek Models' Key Innovative Techniques Submitted 14 March, 2025; originally a
|
||||
2609.15627 DeepSeek -V4-Flash on AMD gfx90a: Correctness Recovery and Inference P Submitted 14 September, 2026; original
|
||||
2503.16304 Bridging Technology and Humanities: Evaluating the Impact of Large Lan Submitted 15 April, 2025; v1 submitted
|
||||
2408.08152 DeepSeek -Prover-V1.5: Harnessing Proof Assistant Feedback for Reinfor Submitted 15 August, 2024; originally
|
||||
2602.14158 A Multi-Agent Framework for Medical AI: Leveraging Fine-Tuned GPT, LLa Submitted 15 February, 2026; originall
|
||||
2504.07128 DeepSeek -R1 Thoughtology: Let's think about LLM Reasoning Submitted 15 January, 2026; v1 submitt
|
||||
2501.08156 Are DeepSeek R1 And Other Reasoning Models More Faithful? Submitted 15 July, 2025; v1 submitted
|
||||
2505.00551 100 Days After DeepSeek -R1: A Survey on Replication Studies and More Submitted 15 May, 2025; v1 submitted 1
|
||||
2502.11164 Quantifying the Capability Boundary of DeepSeek Models: An Application Submitted 15 May, 2025; v1 submitted 1
|
||||
2609.18038 CoAtNet-DeepMoE: A Convolution-Attention Hybrid with DeepSeek Mixture- Submitted 15 September, 2026; original
|
||||
2502.11347 Evaluating the Performance of the DeepSeek Model in Confidential Compu Submitted 16 February, 2025; originall
|
||||
2502.11108 Knowledge Graph-Driven Retrieval-Augmented Generation: Integrating Dee Submitted 16 February, 2025; originall
|
||||
2502.11096 Mixture of Tunable Experts -- Behavior Modification of DeepSeek -R1 at Submitted 16 February, 2025; originall
|
||||
2601.11464 MHA2MLA-VLM: Enabling DeepSeek's Economical Multi-Head Latent Attentio Submitted 16 January, 2026; originally
|
||||
2506.14104 Innovating China's Intangible Cultural Heritage with DeepSeek + MidJou Submitted 16 June, 2025; originally an
|
||||
2503.13549 A Showdown of ChatGPT vs DeepSeek in Solving Programming Tasks Submitted 16 March, 2025; originally a
|
||||
2503.16529 Safety Evaluation and Enhancement of DeepSeek Models in Chinese Contex Submitted 16 May, 2025; v1 submitted 1
|
||||
2510.15009 Can generative AI figure out figurative language? The influence of idi Submitted 16 October, 2025; originally
|
||||
2509.19346 Benchmarking ChatGPT and DeepSeek in April 2025: A Novel Dual Perspect Submitted 16 September, 2025; original
|
||||
2406.11931 DeepSeek -Coder-V2: Breaking the Barrier of Closed-Source Models in Co Submitted 17 June, 2024; originally an
|
||||
2609.19969 DeepSeek -V4.1-Flash: Pushing the Limits of KV Cache Compression Submitted 17 September, 2026; original
|
||||
2608.16393 Security Assessment of DeepSeek Harness with A.I.G: Evaluating Resista Submitted 18 August, 2026; v1 submitte
|
||||
2412.19437 DeepSeek -V3 Technical Report Submitted 18 February, 2025; v1 submit
|
||||
2504.21801 DeepSeek -Prover-V2: Advancing Formal Mathematical Reasoning via Reinf Submitted 18 July, 2025; v1 submitted
|
||||
2505.12625 R1dacted: Investigating Local Censorship in DeepSeek's R1 Language Mod Submitted 18 May, 2025; originally ann
|
||||
2607.20145 SLAI T-Rex: Full-Parameter Post-training of the DeepSeek -V4 Family on Submitted 19 August, 2026; v1 submitte
|
||||
2502.12825 Reasoning and the Trusting Behavior of DeepSeek and GPT: An Experiment Submitted 19 February, 2025; v1 submit
|
||||
2405.04434 DeepSeek -V2: A Strong, Economical, and Efficient Mixture-of-Experts L Submitted 19 June, 2024; v1 submitted
|
||||
2503.10486 LLMs in Disease Diagnosis: A Comparative Study of DeepSeek -R1 and O3 Submitted 19 June, 2025; v1 submitted
|
||||
2503.15092 Towards Understanding the Safety Boundaries of DeepSeek Models: Evalua Submitted 19 March, 2025; originally a
|
||||
2504.02888 A Status Quo Investigation of Large Language Models towards Cost-Effec Submitted 2 April, 2025; originally an
|
||||
2506.21562 FloorPlan- DeepSeek (FPDS): A multimodal approach to floorplan generat Submitted 2 August, 2025; v1 submitted
|
||||
2512.02556 DeepSeek -V3.2: Pushing the Frontier of Open Large Language Models Submitted 2 December, 2025; originally
|
||||
2506.01814 Analysis of LLM Bias (Chinese Propaganda & Anti-US Sentiment) in DeepS Submitted 2 June, 2025; originally ann
|
||||
2503.01923 Output Length Effect on DeepSeek -R1's Safety in Forced Thinking Submitted 2 March, 2025; originally an
|
||||
2502.17487 User Intent to Use DeepSeek for Healthcare Purposes and their Trust in Submitted 2 March, 2025; v1 submitted
|
||||
2509.02036 DeepSeek performs better than other Large Language Models in Dental Ca Submitted 2 September, 2025; originall
|
||||
2502.05769 Digital Twin Buildings: 3D Modeling, GIS Integration, and Visual Descr Submitted 20 April, 2025; v1 submitted
|
||||
2604.03476 Fine-tuning DeepSeek -OCR-2 for Molecular Structure Recognition Submitted 20 April, 2026; v1 submitted
|
||||
2606.09079 FlashMemory- DeepSeek -V4: Lightning Index Ultra-Long Context via Look Submitted 20 July, 2026; v1 submitted
|
||||
2510.18234 DeepSeek -OCR: Contexts Optical Compression Submitted 20 October, 2025; originally
|
||||
2503.09956 DeepSeek -Inspired Exploration of RL-based LLMs and Synergy with Wirel Submitted 20 October, 2025; v1 submitt
|
||||
2502.18505 Comprehensive Analysis of Transparency and Accessibility of ChatGPT, D Submitted 21 February, 2025; originall
|
||||
2503.11068 DeepSeek Powered Solid Dosage Formulation Design and Development Submitted 21 March, 2025; v1 submitted
|
||||
2505.17112 Cultural Value Alignment in Large Language Models: A Prompt-based Anal Submitted 21 May, 2025; originally ann
|
||||
2605.00392 RTPrune: Reading-Twice Inspired Token Pruning for Efficient DeepSeek - Submitted 21 May, 2026; v1 submitted 1
|
||||
2504.16027 Benchmarking LLM for Code Smells Detection: OpenAI GPT-4.0 vs DeepSeek Submitted 22 April, 2025; originally a
|
||||
2505.09343 Insights into DeepSeek -V3: Scaling Challenges and Reflections on Hard Submitted 22 December, 2025; v1 submit
|
||||
2502.18513 Analyzing User Perceptions of Large Language Models (LLMs) on Reddit: Submitted 22 February, 2025; originall
|
||||
2502.16428 Visual Reasoning Evaluation of Grok, Deepseek Janus, Gemini, Qwen, Mis Submitted 22 February, 2025; originall
|
||||
2505.00025 A Method for the Architecture of a Medical Vertical Large Language Mod Submitted 22 July, 2025; v1 submitted
|
||||
2512.20344 A DeepSeek -Powered AI System for Automated Chest Radiograph Interpret Submitted 23 December, 2025; originall
|
||||
2603.26673 Can AI be a Teaching Partner? Evaluating ChatGPT, Gemini, and DeepSeek Submitted 23 February, 2026; originall
|
||||
2507.17944 Evaluating the Performance of AI Text Detectors, Few-Shot and Chain-of Submitted 23 July, 2025; originally an
|
||||
2405.14333 DeepSeek -Prover: Advancing Theorem Proving in LLMs through Large-Scal Submitted 23 May, 2024; originally ann
|
||||
2505.18059 Assessing the performance of 8 AI chatbots in bibliographic reference Submitted 23 May, 2025; originally ann
|
||||
2507.08621 A comprehensive study of LLM-based argument classification: from LLAMA Submitted 24 July, 2025; v1 submitted
|
||||
2511.19333 Learning to Reason: Training LLMs with GPT-OSS or DeepSeek R1 Reasonin Submitted 24 November, 2025; originall
|
||||
2502.03688 A Comparison of DeepSeek and Other LLMs Submitted 25 December, 2025; v1 submit
|
||||
2502.18210 From ChatGPT to DeepSeek : Can LLMs Simulate Humanity? Submitted 25 February, 2025; originall
|
||||
2502.17947 DeepSeek -R1 Outperforms Gemini 2.0 Pro, OpenAI o1, and o3-mini in Bil Submitted 25 February, 2025; originall
|
||||
2503.04783 Comparative Analysis Based on DeepSeek , ChatGPT, and Google Gemini: F Submitted 25 February, 2025; originall
|
||||
2603.19253 A comprehensive study of LLM-based argument classification: from Llama Submitted 25 February, 2026; originall
|
||||
2506.18543 SoK: A Comprehensive Security Analysis of Jailbreak Resilience in GPT Submitted 25 May, 2026; v1 submitted 2
|
||||
2606.19348 DeepSeek -V4: Towards Highly Efficient Million-Token Context Intellige Submitted 26 April, 2026; originally a
|
||||
2502.12893 H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Ja Submitted 26 February, 2025; v1 submit
|
||||
2401.14196 DeepSeek -Coder: When the Large Language Model Meets Programming -- Th Submitted 26 January, 2024; v1 submitt
|
||||
2502.16732 DeepSeek reshaping healthcare in China's tertiary hospitals Submitted 27 February, 2025; v1 submit
|
||||
2504.00016 Medical Reasoning in LLMs: An In-Depth Analysis of DeepSeek R1 Submitted 27 March, 2025; originally a
|
||||
2504.03665 LLM & HPC:Benchmarking DeepSeek's Performance in High-Performance Comp Submitted 28 April, 2025; v1 submitted
|
||||
2505.17702 Seek-CAD: A Self-refined Generative Modeling for 3D Parametric CAD Usi Submitted 28 February, 2026; v1 submit
|
||||
2501.17030 Challenges in Ensuring AI Safety in DeepSeek -R1 Models: The Shortcomi Submitted 28 January, 2025; originally
|
||||
2601.20552 DeepSeek -OCR 2: Visual Causal Flow Submitted 28 January, 2026; originally
|
||||
2509.23676 From Reasoning to Answer: Empirical, Attention-Based and Mechanistic I Submitted 28 September, 2025; original
|
||||
2508.21377 Challenges and Applications of Large Language Models: A Comparison of Submitted 29 August, 2025; originally
|
||||
2506.23128 Are Large Language Models Capable of Deep Relational Reasoning? Insigh Submitted 29 June, 2025; originally an
|
||||
2503.23130 Can DeepSeek Reason Like a Surgeon? An Empirical Evaluation for Vision Submitted 3 April, 2025; v1 submitted
|
||||
2502.01225 The dark deep side of DeepSeek : Fine-tuning attacks against the safet Submitted 3 February, 2025; originally
|
||||
2503.11655 Explainable Sentiment Analysis with DeepSeek -R1: Performance, Efficie Submitted 3 February, 2026; v1 submitt
|
||||
2501.12948 DeepSeek -R1: Incentivizing Reasoning Capability in LLMs via Reinforce Submitted 3 January, 2026; v1 submitte
|
||||
2506.02523 Hardware-Centric Analysis of DeepSeek's Multi-Head Latent Attention Submitted 3 June, 2025; originally ann
|
||||
2503.02032 Comparative Analysis of OpenAI GPT-4o and DeepSeek R1 for Scientific T Submitted 3 March, 2025; originally an
|
||||
2502.14837 Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent At Submitted 3 October, 2025; v1 submitte
|
||||
2502.18467 ChatGPT vs. DeepSeek : A Comparative Study on AI-Based Code Generation Submitted 30 January, 2025; originally
|
||||
2501.18576 Token-Hungry, Yet Precise: DeepSeek R1 Highlights the Need for Multi-S Submitted 30 January, 2025; originally
|
||||
2506.00312 An evaluation of LLMs for generating movie reviews: GPT-4o, Gemini-2.0 Submitted 30 May, 2025; originally ann
|
||||
2504.08120 DeepSeek -R1 vs. o3-mini: How Well can Reasoning LLMs Evaluate MT and Submitted 30 May, 2025; v1 submitted 1
|
||||
2608.30877 Deploying DeepSeek 175B Locally on a Single Consumer-Grade RTX 4060 La Submitted 31 August, 2026; originally
|
||||
2501.18438 o3-mini vs DeepSeek -R1: Which One is Safer? Submitted 31 January, 2025; v1 submitt
|
||||
2507.19493 From Bench to Bedside: A DeepSeek -Powered AI System for Automated Che Submitted 31 May, 2025; originally ann
|
||||
2606.01228 Institutional Trust and the Domestic AI Advantage: Evidence from DeepS Submitted 31 May, 2026; originally ann
|
||||
2504.03219 From ChatGPT to DeepSeek AI: A Comprehensive Analysis of Evolution, De Submitted 4 April, 2025; originally an
|
||||
2401.02954 DeepSeek LLM: Scaling Open-Source Language Models with Longtermism Submitted 5 January, 2024; originally
|
||||
2506.18501 Comparative Evaluation of ChatGPT and DeepSeek Across Key NLP Tasks: S Submitted 5 July, 2025; v1 submitted 2
|
||||
2606.07410 A Comprehensive Anatomy of Human and DeepSeek -R1 LLM Mathematical Rea Submitted 5 June, 2026; originally ann
|
||||
2506.05921 Multi-Modal Large Models Based Beam Prediction: An Example Empowered b Submitted 6 June, 2025; originally ann
|
||||
2609.07008 RedKnot-MLA: Multi-Head Offline-Online Reuse for DeepSeek -V4 Long-Con Submitted 6 September, 2026; originall
|
||||
2502.14926 DeepSeek -V3, GPT-4, Phi-4, and LLaMA-3.3 generate correct code for Lo Submitted 7 April, 2025; v1 submitted
|
||||
2502.02523 Brief analysis of DeepSeek R1 and its implications for Generative AI Submitted 7 February, 2025; v1 submitt
|
||||
2502.11137 Safety Evaluation of DeepSeek Models in Chinese Contexts Submitted 7 May, 2025; v1 submitted 16
|
||||
2601.03714 Visual Merit or Linguistic Crutch? A Close Look at DeepSeek -OCR Submitted 8 January, 2026; v1 submitte
|
||||
2502.17764 DeepSeek vs. ChatGPT vs. Claude: A Comparative Study for Scientific Co Submitted 8 March, 2025; v1 submitted
|
||||
2605.08615 DSPE: An Energy-Efficient Edge Processor for DeepSeek Inference with M Submitted 8 May, 2026; originally anno
|
||||
2606.10392 Instruction Finetuning DeepSeek -R1-8B Model Using LoRA and NEFTune Submitted 9 June, 2026; originally ann
|
||||
@@ -0,0 +1,80 @@
|
||||
# 论文中文精读版写作规范
|
||||
|
||||
你是学术论文中文精读助手。请为一篇 DeepSeek 官方论文撰写**中文精读版 Markdown**。
|
||||
|
||||
## 输入
|
||||
|
||||
- **摘要文件**：`D:\onedrive\note_sync\大模型\deepseek\_digest\<arXivID>.txt`
|
||||
含标题、作者、章节标题清单、正文前半部分与结尾部分。**先读这个**。
|
||||
- **完整正文**：`D:\onedrive\note_sync\大模型\deepseek\_text\<arXivID>.txt`
|
||||
如摘要文件信息不足（例如需要具体实验数据、某章节细节），用 read 工具的 `offset` / `limit`
|
||||
分段读取该文件的相关部分。正文中含有章节标题（以 `## ` 开头）可作为定位锚点。
|
||||
|
||||
严禁凭空编造。所有数字、结论必须能在上述文件中找到依据；找不到就写「论文未明确说明」。
|
||||
|
||||
## 输出
|
||||
|
||||
用 write 工具写入（UTF-8），路径由调用方在任务中指定，形如：
|
||||
`D:\onedrive\note_sync\大模型\deepseek\<类别目录>\<文件名前缀>.md`
|
||||
|
||||
## 结构（必须严格遵循）
|
||||
|
||||
````markdown
|
||||
# <中文标题>
|
||||
|
||||
> **原文标题**：<英文原标题>
|
||||
> **arXiv**：[<arXivID>](https://arxiv.org/abs/<arXivID>) ｜ **发布**：<日期>
|
||||
> **作者/机构**：<第一作者等 + DeepSeek（深度求索）>
|
||||
> **原文 PDF**：[<文件名前缀>.pdf](./<文件名前缀>.pdf)
|
||||
|
||||
## 一、摘要（全文翻译）
|
||||
|
||||
<把论文 Abstract 逐句完整翻译成中文。不要删减、不要概括、不要跳跃。>
|
||||
|
||||
## 二、核心贡献
|
||||
|
||||
1. **<要点小标题>**：一句话说清「提出了什么、解决了什么问题、效果如何」。
|
||||
2. …（3–6 条）
|
||||
|
||||
## 三、方法与技术要点
|
||||
|
||||
### 3.1 <小节名>
|
||||
<说明关键设计。公式用 LaTeX 保留原样，例如 $\mathcal{L} = -\sum_i \log p_i$。>
|
||||
|
||||
### 3.2 <小节名>
|
||||
…
|
||||
|
||||
## 四、主要实验结果
|
||||
|
||||
| 基准 / 任务 | 本模型 | 对比对象 | 说明 |
|
||||
|---|---|---|---|
|
||||
| … | … | … | … |
|
||||
|
||||
<表格下方 2–4 句话总结：哪些任务领先、领先幅度、在哪些方面仍有差距。数字保留原始精度。>
|
||||
|
||||
## 五、结论、局限与影响
|
||||
|
||||
- **结论**：…
|
||||
- **局限性**：<论文中作者自述的局限；若未提及则写"论文未明确说明">
|
||||
- **影响**：…
|
||||
|
||||
## 六、关键术语对照表
|
||||
|
||||
| English | 中文 |
|
||||
|---|---|
|
||||
| … | … |
|
||||
|
||||
（8–15 条，覆盖该论文最核心的专有名词）
|
||||
````
|
||||
|
||||
## 写作要求
|
||||
|
||||
- 全文使用中文；专有名词首次出现时保留英文原词，例如「多头潜在注意力（Multi-Head Latent Attention, MLA）」。
|
||||
- 篇幅：正文 **1200–2500 汉字**（表格、公式、术语表不计入）。宁可信息密度高，不要注水。
|
||||
- 保留论文的章节逻辑，不要重新编排论证顺序。
|
||||
- 若论文含重要公式、算法伪代码或架构图说明，在「方法与技术要点」中用文字+公式复述其含义。
|
||||
- 讲清「为什么这么做」：不仅罗列设计，还要说明该设计针对的痛点。
|
||||
|
||||
## 完成后
|
||||
|
||||
只回复一行：`OK <输出文件名> <汉字数>`。不要输出论文正文到对话中。
|
||||
@@ -0,0 +1,6 @@
|
||||
https://huggingface.co/deepseek-ai
|
||||
https://github.com/deepseek-ai
|
||||
https://huggingface.co/papers?q=DeepSeek
|
||||
https://deepseekai.guide/news/deepseek-research-papers/
|
||||
https://hub.baai.ac.cn/view/54241
|
||||
https://www.deepseek.com/
|
||||
@@ -0,0 +1,4 @@
|
||||
https://www.deepseek.com/research
|
||||
https://www.deepseek.com/news
|
||||
https://www.deepseek.com/en/research
|
||||
https://api-docs.deepseek.com/news/news250325
|
||||
@@ -0,0 +1,5 @@
|
||||
https://www.deepseek.com/research/
|
||||
https://www.deepseek.com/research-index/
|
||||
https://www.deepseek.com/news/research/
|
||||
https://www.deepseek.com/en/news/
|
||||
https://www.deepseek.com/publications/
|
||||
@@ -0,0 +1,3 @@
|
||||
https://www.deepseek.com/transparency/
|
||||
https://github.com/deepseek-ai/LPLB
|
||||
https://github.com/intentlink/deepseek-news-research/blob/main/2025/linear-programming-load-balancer.md
|
||||
@@ -0,0 +1 @@
|
||||
https://huggingface.co/collections/Presidentlin/deepseek-papers-674c536aa6acddd9bc98c2ac
|
||||
File diff suppressed because one or more lines are too long
@@ -0,0 +1,28 @@
|
||||
[
|
||||
{
|
||||
"id": "2407.01906",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models",
|
||||
"authors": "Zihan Wang , Deli Chen , Damai Dai , Runxin Xu , Zhuoshu Li , Y. Wu",
|
||||
"abstract": "Parameter-efficient fine-tuning (PEFT) is crucial for customizing Large Language Models (LLMs) with constrained resources. Although there have been various PEFT methods for dense-architecture LLMs, PEFT for sparse-architecture LLMs is still underexplored. In this work, we study the PEFT method for LLMs with the Mixture-of-Experts (MoE) architecture and the contents of this work are mainly threefold: (1) We investigate the dispersion degree of the activated experts in customized tasks, and found that the routing distribution for a specific task tends to be highly concentrated, while the distribution of activated experts varies significantly across different tasks. (2) We propose Expert-Specialized Fine-Tuning, or ESFT, which tunes the experts most relevant to downstream tasks while freezing the other experts and modules; experimental results demonstrate that our method not only improves the tuning efficiency, but also matches or even surpasses the performance of full-parameter fine-tuning. (3) We further analyze the impact of the MoE architecture on expert-specialized fine-tuning. We find that MoE models with finer-grained experts are more advantageous in selecting the combination of experts that are most relevant to downstream tasks, thereby enhancing both the training efficiency and effectiveness. Our code is available at this https URL .",
|
||||
"subjects": "Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG)",
|
||||
"dateline": "[Submitted on 2 Jul 2024 ( v1 ), last revised 5 Jul 2024 (this version, v2)]",
|
||||
"comments": "",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
},
|
||||
{
|
||||
"id": "2504.02495",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "Inference-Time Scaling for Generalist Reward Modeling",
|
||||
"authors": "Zijun Liu , Peiyi Wang , Runxin Xu , Shirong Ma , Chong Ruan , Peng Li , Yang Liu , Yu Wu",
|
||||
"abstract": "Reinforcement learning (RL) has been widely adopted in post-training for large language models (LLMs) at scale. Recently, the incentivization of reasoning capabilities in LLMs from RL indicates that $\\textit{proper learning methods could enable effective inference-time scalability}$. A key challenge of RL is to obtain accurate reward signals for LLMs in various domains beyond verifiable questions or artificial rules. In this work, we investigate how to improve reward modeling (RM) with more inference compute for general queries, i.e. the $\\textbf{inference-time scalability of generalist RM}$, and further, how to improve the effectiveness of performance-compute scaling with proper learning methods. For the RM approach, we adopt pointwise generative reward modeling (GRM) to enable flexibility for different input types and potential for inference-time scaling. For the learning method, we propose Self-Principled Critique Tuning (SPCT) to foster scalable reward generation behaviors in GRMs through online RL, to generate principles adaptively and critiques accurately, resulting in $\\textbf{DeepSeek-GRM}$ models. Furthermore, for effective inference-time scaling, we use parallel sampling to expand compute usage, and introduce a meta RM to guide voting process for better scaling performance. Empirically, we show that SPCT significantly improves the quality and scalability of GRMs, outperforming existing methods and models in various RM benchmarks without severe biases, and could achieve better performance compared to training-time scaling. DeepSeek-GRM still meets challenges in some tasks, which we believe can be addressed by future efforts in generalist reward systems. The models are released at Hugging Face and ModelScope.",
|
||||
"subjects": "Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG)",
|
||||
"dateline": "[Submitted on 3 Apr 2025 ( v1 ), last revised 25 Sep 2025 (this version, v3)]",
|
||||
"comments": "Preprint, under review. 44 pages. Models are available at this https URL and this https URL",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,15 @@
|
||||
[
|
||||
{
|
||||
"id": "2512.24880",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "mHC: Manifold-Constrained Hyper-Connections",
|
||||
"authors": "Zhenda Xie , Yixuan Wei , Huanqi Cao , Chenggang Zhao , Chengqi Deng , Jiashi Li , Damai Dai , Huazuo Gao , Jiang Chang , Kuai Yu , Liang Zhao , Shangyan Zhou , Zhean Xu , Zhengyan Zhang , Wangding Zeng , Shengding Hu , Yuqing Wang , Jingyang Yuan , Lean Wang , Wenfeng Liang",
|
||||
"abstract": "Recently, studies exemplified by Hyper-Connections (HC) have extended the ubiquitous residual connection paradigm established over the past decade by expanding the residual stream width and diversifying connectivity patterns. While yielding substantial performance gains, this diversification fundamentally compromises the identity mapping property intrinsic to the residual connection, which causes severe training instability and restricted scalability, and additionally incurs notable memory access overhead. To address these challenges, we propose Manifold-Constrained Hyper-Connections (mHC), a general framework that projects the residual connection space of HC onto a specific manifold to restore the identity mapping property, while incorporating rigorous infrastructure optimization to ensure efficiency. Empirical experiments demonstrate that mHC is effective for training at scale, offering tangible performance improvements and superior scalability. We anticipate that mHC, as a flexible and practical extension of HC, will contribute to a deeper understanding of topological architecture design and suggest promising directions for the evolution of foundational models.",
|
||||
"subjects": "Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG)",
|
||||
"dateline": "[Submitted on 31 Dec 2025 ( v1 ), last revised 5 Jan 2026 (this version, v2)]",
|
||||
"comments": "",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,15 @@
|
||||
[
|
||||
{
|
||||
"id": "2602.21548",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference",
|
||||
"authors": "Yongtong Wu , Shaoyuan Chen , Yinmin Zhong , Rilin Huang , Yixuan Tan , Wentao Zhang , Liyue Zhang , Shangyan Zhou , Yuxuan Liu , Shunfeng Zhou , Mingxing Zhang , Xin Jin , Panpan Huang",
|
||||
"abstract": "The performance of multi-turn, agentic LLM inference is increasingly dominated by KV-Cache storage I/O rather than computation. In prevalent disaggregated architectures, loading the massive KV-Cache from external storage creates a fundamental imbalance: storage NICs on prefill engines become bandwidth-saturated, while those on decoding engines remain idle. This asymmetry severely constrains overall system throughput. We present DualPath, an inference system that breaks this bottleneck by introducing dual-path KV-Cache loading. Beyond the traditional storage-to-prefill path, DualPath enables a novel storage-to-decode path, in which the KV-Cache is loaded into decoding engines and then efficiently transferred to prefill engines via RDMA over the compute network. DualPath combines this optimized data path -- which inherently avoids network congestion and avoids interference with latency-critical model execution communications -- with a global scheduler that dynamically balances load across prefill and decode engines. Our evaluation on three models with production agentic workloads demonstrates that DualPath improves offline inference throughput by up to 1.87$\\times$ on our in-house inference system. It can also improve online serving throughput by an average factor of 1.96$\\times$ without violating SLO.",
|
||||
"subjects": "Distributed, Parallel, and Cluster Computing (cs.DC)",
|
||||
"dateline": "[Submitted on 25 Feb 2026 ( v1 ), last revised 26 Feb 2026 (this version, v2)]",
|
||||
"comments": "",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,67 @@
|
||||
[
|
||||
{
|
||||
"id": "2310.16818",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "DreamCraft3D: Hierarchical 3D Generation with Bootstrapped Diffusion Prior",
|
||||
"authors": "Jingxiang Sun , Bo Zhang , Ruizhi Shao , Lizhen Wang , Wen Liu , Zhenda Xie , Yebin Liu",
|
||||
"abstract": "We present DreamCraft3D, a hierarchical 3D content generation method that produces high-fidelity and coherent 3D objects. We tackle the problem by leveraging a 2D reference image to guide the stages of geometry sculpting and texture boosting. A central focus of this work is to address the consistency issue that existing works encounter. To sculpt geometries that render coherently, we perform score distillation sampling via a view-dependent diffusion model. This 3D prior, alongside several training strategies, prioritizes the geometry consistency but compromises the texture fidelity. We further propose Bootstrapped Score Distillation to specifically boost the texture. We train a personalized diffusion model, Dreambooth, on the augmented renderings of the scene, imbuing it with 3D knowledge of the scene being optimized. The score distillation from this 3D-aware diffusion prior provides view-consistent guidance for the scene. Notably, through an alternating optimization of the diffusion prior and 3D scene representation, we achieve mutually reinforcing improvements: the optimized 3D scene aids in training the scene-specific diffusion model, which offers increasingly view-consistent guidance for 3D optimization. The optimization is thus bootstrapped and leads to substantial texture boosting. With tailored 3D priors throughout the hierarchical generation, DreamCraft3D generates coherent 3D objects with photorealistic renderings, advancing the state-of-the-art in 3D content generation. Code available at this https URL .",
|
||||
"subjects": "Computer Vision and Pattern Recognition (cs.CV) ; Computational Geometry (cs.CG)",
|
||||
"dateline": "[Submitted on 25 Oct 2023 ( v1 ), last revised 26 Oct 2023 (this version, v2)]",
|
||||
"comments": "Project Page: this https URL",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
},
|
||||
{
|
||||
"id": "2408.15664",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts",
|
||||
"authors": "Lean Wang , Huazuo Gao , Chenggang Zhao , Xu Sun , Damai Dai",
|
||||
"abstract": "For Mixture-of-Experts (MoE) models, an unbalanced expert load will lead to routing collapse or increased computational overhead. Existing methods commonly employ an auxiliary loss to encourage load balance, but a large auxiliary loss will introduce non-negligible interference gradients into training and thus impair the model performance. In order to control load balance while not producing undesired gradients during training, we propose Loss-Free Balancing, featured by an auxiliary-loss-free load balancing strategy. To be specific, before the top-K routing decision, Loss-Free Balancing will first apply an expert-wise bias to the routing scores of each expert. By dynamically updating the bias of each expert according to its recent load, Loss-Free Balancing can consistently maintain a balanced distribution of expert load. In addition, since Loss-Free Balancing does not produce any interference gradients, it also elevates the upper bound of model performance gained from MoE training. We validate the performance of Loss-Free Balancing on MoE models with up to 3B parameters trained on up to 200B tokens. Experimental results show that Loss-Free Balancing achieves both better performance and better load balance compared with traditional auxiliary-loss-controlled load balancing strategies.",
|
||||
"subjects": "Machine Learning (cs.LG) ; Computation and Language (cs.CL)",
|
||||
"dateline": "[Submitted on 28 Aug 2024]",
|
||||
"comments": "",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
},
|
||||
{
|
||||
"id": "2411.07975",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation",
|
||||
"authors": "Yiyang Ma , Xingchao Liu , Xiaokang Chen , Wen Liu , Chengyue Wu , Zhiyu Wu , Zizheng Pan , Zhenda Xie , Haowei Zhang , Xingkai yu , Liang Zhao , Yisong Wang , Jiaying Liu , Chong Ruan",
|
||||
"abstract": "We present JanusFlow, a powerful framework that unifies image understanding and generation in a single model. JanusFlow introduces a minimalist architecture that integrates autoregressive language models with rectified flow, a state-of-the-art method in generative modeling. Our key finding demonstrates that rectified flow can be straightforwardly trained within the large language model framework, eliminating the need for complex architectural modifications. To further improve the performance of our unified model, we adopt two key strategies: (i) decoupling the understanding and generation encoders, and (ii) aligning their representations during unified training. Extensive experiments show that JanusFlow achieves comparable or superior performance to specialized models in their respective domains, while significantly outperforming existing unified approaches across standard benchmarks. This work represents a step toward more efficient and versatile vision-language models.",
|
||||
"subjects": "Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL)",
|
||||
"dateline": "[Submitted on 12 Nov 2024 ( v1 ), last revised 24 Mar 2025 (this version, v2)]",
|
||||
"comments": "Accepted by CVPR 2025",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
},
|
||||
{
|
||||
"id": "2502.07316",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "CodeI/O: Condensing Reasoning Patterns via Code Input-Output Prediction",
|
||||
"authors": "Junlong Li , Daya Guo , Dejian Yang , Runxin Xu , Yu Wu , Junxian He",
|
||||
"abstract": "Reasoning is a fundamental capability of Large Language Models. While prior research predominantly focuses on enhancing narrow skills like math or code generation, improving performance on many other reasoning tasks remains challenging due to sparse and fragmented training data. To address this issue, we propose CodeI/O, a novel approach that systematically condenses diverse reasoning patterns inherently embedded in contextually-grounded codes, through transforming the original code into a code input-output prediction format. By training models to predict inputs/outputs given code and test cases entirely in natural language as Chain-of-Thought (CoT) rationales, we expose them to universal reasoning primitives -- like logic flow planning, state-space searching, decision tree traversal, and modular decomposition -- while decoupling structured reasoning from code-specific syntax and preserving procedural rigor. Experimental results demonstrate CodeI/O leads to consistent improvements across symbolic, scientific, logic, math & numerical, and commonsense reasoning tasks. By matching the existing ground-truth outputs or re-executing the code with predicted inputs, we can verify each prediction and further enhance the CoTs through multi-turn revision, resulting in CodeI/O++ and achieving higher performance. Our data and models are available at this https URL .",
|
||||
"subjects": "Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI)",
|
||||
"dateline": "[Submitted on 11 Feb 2025 ( v1 ), last revised 21 May 2025 (this version, v4)]",
|
||||
"comments": "ICML 2025",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
},
|
||||
{
|
||||
"id": "2502.11089",
|
||||
"ok": true,
|
||||
"version": "1",
|
||||
"title": "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention",
|
||||
"authors": "Jingyang Yuan , Huazuo Gao , Damai Dai , Junyu Luo , Liang Zhao , Zhengyan Zhang , Zhenda Xie , Y. X. Wei , Lean Wang , Zhiping Xiao , Yuqing Wang , Chong Ruan , Ming Zhang , Wenfeng Liang , Wangding Zeng",
|
||||
"abstract": "Long-context modeling is crucial for next-generation language models, yet the high computational cost of standard attention mechanisms poses significant computational challenges. Sparse attention offers a promising direction for improving efficiency while maintaining model capabilities. We present NSA, a Natively trainable Sparse Attention mechanism that integrates algorithmic innovations with hardware-aligned optimizations to achieve efficient long-context modeling. NSA employs a dynamic hierarchical sparse strategy, combining coarse-grained token compression with fine-grained token selection to preserve both global context awareness and local precision. Our approach advances sparse attention design with two key innovations: (1) We achieve substantial speedups through arithmetic intensity-balanced algorithm design, with implementation optimizations for modern hardware. (2) We enable end-to-end training, reducing pretraining computation without sacrificing model performance. As shown in Figure 1, experiments show the model pretrained with NSA maintains or exceeds Full Attention models across general benchmarks, long-context tasks, and instruction-based reasoning. Meanwhile, NSA achieves substantial speedups over Full Attention on 64k-length sequences across decoding, forward propagation, and backward propagation, validating its efficiency throughout the model lifecycle.",
|
||||
"subjects": "Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG)",
|
||||
"dateline": "[Submitted on 16 Feb 2025 ( v1 ), last revised 27 Feb 2025 (this version, v2)]",
|
||||
"comments": "",
|
||||
"journal_ref": "",
|
||||
"doi": ""
|
||||
}
|
||||
]
|
||||
@@ -0,0 +1,88 @@
|
||||
"""Verify arXiv IDs by fetching their abs pages and extracting metadata.
|
||||
|
||||
Usage: python verify_ids.py <ids.txt> <out.json>
|
||||
ids.txt: one arXiv id per line (may include comments after '#')
|
||||
"""
|
||||
import urllib.request
|
||||
import re
|
||||
import json
|
||||
import sys
|
||||
import time
|
||||
import html as htmlmod
|
||||
|
||||
UA = 'Mozilla/5.0 (compatible; paperbot/1.0)'
|
||||
|
||||
|
||||
def get(url, timeout=40):
|
||||
req = urllib.request.Request(url, headers={'User-Agent': UA})
|
||||
with urllib.request.urlopen(req, timeout=timeout) as r:
|
||||
return r.status, r.read().decode('utf-8', 'ignore')
|
||||
|
||||
|
||||
def strip(s):
|
||||
s = re.sub(r'<[^>]+>', ' ', s)
|
||||
s = htmlmod.unescape(s)
|
||||
return ' '.join(s.split())
|
||||
|
||||
|
||||
def extract(aid, t):
|
||||
def find(pat, flags=re.S):
|
||||
m = re.search(pat, t, flags)
|
||||
return strip(m.group(1)) if m else ''
|
||||
|
||||
title = find(r'<h1 class="title mathjax">(.*?)</h1>')
|
||||
title = re.sub(r'^Title:\s*', '', title)
|
||||
authors = find(r'<div class="authors">(.*?)</div>')
|
||||
authors = re.sub(r'^Authors:\s*', '', authors)
|
||||
abstract = find(r'<blockquote class="abstract mathjax">(.*?)</blockquote>')
|
||||
abstract = re.sub(r'^Abstract:\s*', '', abstract)
|
||||
subj = find(r'<td class="tablecell subjects">(.*?)</td>')
|
||||
dateline = find(r'<div class="dateline">(.*?)</div>')
|
||||
comments = find(r'<td class="tablecell comments[^"]*">(.*?)</td>')
|
||||
journal = find(r'<td class="tablecell jref">(.*?)</td>')
|
||||
doi = find(r'<td class="tablecell doi">(.*?)</td>')
|
||||
m = re.search(r'\[v(\d+)\]', t)
|
||||
ver = m.group(1) if m else ''
|
||||
return {
|
||||
'id': aid,
|
||||
'version': ver,
|
||||
'title': title,
|
||||
'authors': authors,
|
||||
'abstract': abstract,
|
||||
'subjects': subj,
|
||||
'dateline': dateline,
|
||||
'comments': comments,
|
||||
'journal_ref': journal,
|
||||
'doi': doi,
|
||||
}
|
||||
|
||||
|
||||
def main():
|
||||
ids = []
|
||||
with open(sys.argv[1], encoding='utf-8') as f:
|
||||
for line in f:
|
||||
line = line.split('#')[0].strip()
|
||||
if line:
|
||||
ids.append(line)
|
||||
out = []
|
||||
for aid in ids:
|
||||
rec = {'id': aid, 'ok': False}
|
||||
for attempt in range(3):
|
||||
try:
|
||||
st, t = get('https://arxiv.org/abs/' + aid)
|
||||
if st == 200 and 'Title:' in t:
|
||||
rec.update(extract(aid, t))
|
||||
rec['ok'] = True
|
||||
break
|
||||
except Exception as e: # noqa: BLE001
|
||||
rec['error'] = str(e)[:120]
|
||||
time.sleep(3)
|
||||
print(('OK ' if rec['ok'] else 'MISS'), aid, rec.get('title', '')[:90], file=sys.stderr)
|
||||
out.append(rec)
|
||||
time.sleep(1.5)
|
||||
with open(sys.argv[2], 'w', encoding='utf-8') as f:
|
||||
json.dump(out, f, ensure_ascii=False, indent=1)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -0,0 +1,48 @@
|
||||
"""Verify every planned output exists and looks sane.
|
||||
|
||||
Usage: python verify_outputs.py <plan.json> <base_dir>
|
||||
"""
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
|
||||
plan = json.load(open(sys.argv[1], encoding='utf-8'))
|
||||
base = sys.argv[2]
|
||||
missing = []
|
||||
small = []
|
||||
ok = []
|
||||
for p in plan:
|
||||
md = os.path.join(base, p['md_path'])
|
||||
pdf = os.path.join(base, p['pdf_path'])
|
||||
problems = []
|
||||
if not os.path.exists(pdf) or os.path.getsize(pdf) < 20000:
|
||||
problems.append('pdf')
|
||||
if not os.path.exists(md):
|
||||
problems.append('md-missing')
|
||||
else:
|
||||
t = open(md, encoding='utf-8').read()
|
||||
cjk = len(re.findall(r'[\u4e00-\u9fff]', t))
|
||||
if cjk < 900:
|
||||
problems.append('md-short(%d)' % cjk)
|
||||
for sec in ['摘要', '核心贡献', '方法', '实验', '结论']:
|
||||
if sec not in t:
|
||||
problems.append('missing-section:' + sec)
|
||||
if problems:
|
||||
small.append((p['id'], cjk, problems))
|
||||
else:
|
||||
ok.append((p['id'], cjk))
|
||||
if problems and not any(x[0] == p['id'] for x in small):
|
||||
missing.append((p['id'], problems))
|
||||
|
||||
print('OK: %d / %d' % (len(ok), len(plan)))
|
||||
for i, c in ok:
|
||||
print(' ok %s %d 汉字' % (i, c))
|
||||
if small:
|
||||
print('NEEDS ATTENTION:')
|
||||
for i, c, pr in small:
|
||||
print(' %s %d 汉字 %s' % (i, c, pr))
|
||||
if missing:
|
||||
print('MISSING FILES:')
|
||||
for i, pr in missing:
|
||||
print(' %s %s' % (i, pr))
|
||||
Reference in New Issue
Block a user