How to Spend Your Oracle Budget: Practical Guidance for Protein Structure Prediction Models

arXiv 2609.15973（cs.CL，2026-09-14，Yang/Yin/Wu）提出Discovery Foundation Models (DFM)框架，把发现能力形式化为 7 项耦合能力：问题发现问题形式化表征构建假设形成干预证据驱动的修正持续发现改进；作用于可修订的研究状态实例化为 Zetema（显式研究状态动力学 + 验证/实验 gate + 外部 grounding + 跨任务 Discovery Skill 演化）；GALILEO 是真实治疗发现的闭环系统，Dry-Lab 推理机器人/手动 Wet-Lab 实验外部生物证据迭代假设与设计修订形成物理发现闭环；并给出 capability formation + process-centered evaluation 的统一方法论...

本文研究：Why Does CLAUDE.md Keep Growing? Catastrophic Remembering in Agentic Coding摘要要点：Agentic coding READMEs like CLAUDE.md grow without bound in real repositories, stopping only when the repository retires or someone rewrites the file wholesale....

本文研究：SkillZip摘要要点：Self-evolving agents accumulate reusable skills by appending successful procedures and failure fixes. Over time, the same requirement is often restated in several branches, examples, and warnings, while common action sequences are copied rather than reused. The resulting skill becomes expensive to inject and difficult to maintain....

本文研究：Long-Horizon AI Research for Grothendieck Constant摘要要点：AI agents are increasingly used in mathematics research, but it is often unclear how to use them effectively....