STOC 2027: ~75% of 1,665 submissions report AI involvement in proofs and results
fortnow · x · 2026-11-05
STOC 2027 publicly posted all 1,665 submissions and author identities immediately — the most revealing disclosure of any conference. An independent Submission Atlas built with ChatGPT quantifies AI involvement:
- 91.2% of submissions include some AI-use disclosure; 11 explicitly deny AI/LLM use, 128 have no disclosure located.
- 74.7% report AI involvement in developing proofs or results (supporting arguments, proof repairs, algorithms, constructions); in algorithmic game theory the figure reaches 85.9%.
- Model usage: 61.1% report ChatGPT/GPT but no Claude Opus; 15.6% report both; 1.8% only Opus; Gemini is sparse at 6.5%, mostly alongside other models.
- 24.4% of the 2,988 normalized author names (about 730) have no earlier indexed paper through end of 2025.
The most direct quantitative snapshot yet of how deeply AI is embedded in top-tier theory research.
More from Research
- BioSimBench launches: frontier models partially fail at predicting real cell responses; a 27B RL-trained model beats them — divy93t · 2026-11-05
- Stanford's Joules-to-GDP framework measures how much economic value AI produces per joule of energy — HazyResearch · 2026-11-05
- Prefilled-CoT evals are an upper bound, not a strict fix, for CoT controllability — tomekkorbak · 2026-11-05
- Ed Witten's new paper thanks Claude Opus 5.5 for assistance on many details — inductionheads · 2026-11-05
- Tiny neural nets approximate cloud scattering, hitting 1080p 60fps on iPhone 12 — mflux · 2026-11-05
- Cognition shares mid-training recipe insights: token efficiency plots for data ablations — carlesgelada · 2026-11-05