AI agent resolves 100+ of 244 open math problems; proofs withheld to avoid contamination
wellecks · x · 2026-10-20
A team led by Zeyu Zheng released results on a benchmark of 244 open math problems: the Leanimum Agent resolved over 100 of them. Paper, benchmark, and the Leanimum Agent are public, but the team says proofs won't be released — keeping benchmark solutions private is good practice to mitigate training-data contamination. A notable empirical advance for AI in formal mathematics (Lean).
More from Research
- OR-License benchmark finds LLM agents fall short of OR experts — HKUST-DSAIL · 2026-10-20
- Fréchet Forcing trains long-horizon world models from short clips — Yian Zhao · 2026-10-20
- SERA trains recursive agents to self-generate evaluation rubrics — TongjiUniversity · 2026-10-20
- MaSS makes self-supervised depth estimation scalable via scene structure — Tongji · 2026-10-20
- CoPE boosts agent RL sample efficiency via multi-step action forecasting — Haofei Yu · 2026-10-20
- RSIFlow: evolving meta-agent for adaptive self-improvement workflows — Beihang · 2026-10-20