Épi 自我改进研究智能体：刷新 120 项记录并证明 Tuza 猜想部分情形
一个能跨会话持续做研究的智能体，数学上刷了 120 项记录，写求解器还比 Codex 便宜 36% 且更强。
Épi 是一个能跨多次模型会话持续研究同一问题的自我改进研究智能体，已创下 120 项记录，包括将 n-queens 常数的区间收窄到原先的约 1/84。它通过持久研究历史、分层记忆、可回退的版本分支和独立审稿机制运作，并用固定 verifier 评分以防止其改动评判标准。其他成果包括证明 Cohn-Elkies 界在维度 10 和 11 无法取等，以及 Tuza 猜想在 split graphs 一个情形的计算机辅助证明。轻量版 Épi-turbo 专写路由与调度求解器，在 30 个 FrontierOR 任务上比 Codex 高 5.4%，成本仅为其 64%。
Self-improving research agents are one of the most discussed ideas in AI right now.
(bookmark it)
Épi is a research agent that keeps working on the same research problem across many model sessions. It set 120 records, including an interval for the n-queens constant about 84 times narrower than before.
It has a persistent research history, hierarchical memory, versioned branches it can return to, and an independent reviewer. A fixed verifier scores every output, so the agent cannot change how it is judged.
Other results include exact certificates showing the Cohn-Elkies bound cannot be sharp in dimensions 10 and 11, and a computer-assisted proof of a case of Tuza's conjecture for split graphs.
A lighter version, Épi-turbo, writes solvers for routing and scheduling problems. On 30 FrontierOR tasks, it scores 5.4% higher than Codex at 64% of the model cost, with the same model and reasoning effort.
https://t.co/XlQVBadaff
Chat with Paper: https://t.co/7yKdUBypKT
