GPT-6.1 Sol audited ~800 math papers: 6% had false main theorems, now down to 1.5%
141_1337 · reddit · 2026-10-16
Using GPT-6.1 Sol to audit the main theorems of nearly 800 arXiv math papers, Jasper Dekoninck et al. found 6% of pre-Aug 2026 papers contained false main theorems and another 30% had minor mistakes; researchers estimate 90% of the model's major findings are correct. Notably, as AI adoption surged, the major-error rate in papers from the last two months fell to just 1.5%, suggesting AI-assisted review is improving paper quality.
Related event: GPT-6.1 Finds Major Errors in 6% of Math Papers' Main Theorems(2 posts)→
More from Models
- Gemini 4 Argon hits AssemblyAI's LLM Gateway, tops voice bench at 99.5 across 50+ models — AssemblyAI · 2026-10-16
- Meta's open-weight Muse Spark Open hits Ollama with 1M-token context and Claude Code support — ollama · 2026-10-16
- 'Opus 5.5' collage demos spark claim: models have taste, no moat — yuwen_lu_ · 2026-10-16
- PrismML ships Bonsai 2 ternary models: 9B fits in 2GB for local AI — gajesh · 2026-10-16
- Apple MLX Runs 816B Muse Spark Open at 36+ Tokens/s on Four Mac Studios — andrejusb · 2026-10-16
- vLLM Ships Day-0 Support for Meta's 816B Open-Weight Muse Spark Open — woosuk_k · 2026-10-16