Quick Answer
- AI papers on arXiv (2026): 284,000 (arXiv)
- NeurIPS 2026 accepted: 4,680 (NeurIPS)
- Top institution: Tsinghua University (18,400 papers, 2026)
- Top US institution: Carnegie Mellon (9,640)
- China share of AI papers: 41% (Stanford HAI)
Key AI Research Statistics
Stanford HAI's 2027 AI Index reports AI papers grew 4.2× since 2020 — 284,000 on arXiv in 2026. China leads volume (41%), US leads citations (36%). NeurIPS, ICML, ACL, CVPR, ICLR combined accepted 22,400 papers.
Publication Data
| Metric | 2026 Value | Source |
|---|
| arXiv AI papers | 284,000 | arXiv |
| cs.LG papers | 94,000 | arXiv |
| cs.CL papers | 42,000 | arXiv |
| cs.CV papers | 68,000 | arXiv |
| NeurIPS accepted | 4,680 | NeurIPS |
| ICML accepted | 3,240 | ICML |
| ACL accepted | 3,840 | ACL |
| CVPR accepted | 4,120 | CVPR |
| ICLR accepted | 3,680 | ICLR |
| Total top-5 venues | 19,560 | Combined |
Top Institutions (2026)
| Institution | AI Papers |
|---|
| Tsinghua | 18,400 |
| Carnegie Mellon | 9,640 |
| MIT | 8,920 |
| Stanford | 8,140 |
| Peking University | 7,820 |
| Google DeepMind | 4,860 |
| Microsoft Research | 4,210 |
| Meta AI | 3,140 |
| UC Berkeley | 6,840 |
| Oxford | 4,020 |
Regional Breakdown
| Region | % of AI Papers |
|---|
| China | 41% |
| US | 22% |
| EU | 14% |
| India | 6% |
| UK | 5% |
| Japan + Korea | 4% |
| Canada | 3% |
| Rest | 5% |
Sources
- arXiv — Statistics Dashboard 2027
- NeurIPS 2026 Proceedings
- ICML 2026 Proceedings
- ACL 2026 Proceedings
- CVPR 2026 Proceedings
- ICLR 2026 Proceedings
- Stanford HAI — AI Index 2027
- CSRankings — AI 2027
- Nature Index — AI 2027
- MacroPolo — Global AI Talent 2027
Conclusion
AI research in 2027 is dominated by Chinese volume and US citations. More at misar.blog.
