
从8月11日人工智能栏目筛出36篇直接命中与3篇邻近工作,重点比较失败轨迹蒸馏、技能门控、长程恢复、因果归因与安全回滚。
| 顺序 | 原题 | 重点读什么 | 判定 |
|---|---|---|---|
| 1 | Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled Skills | 失败轨迹怎样被离线蒸馏成可复用技能,且不更新权重 | 精读 |
| 2 | FailForge: Distilling Procedural Competence from Persistent Failures into Code Agents | 持续失败如何转成训练信号并内化到代码智能体 | 精读 |
| 3 | Branch2Skill: Efficient Skill Evolution Through Reasoning Trees | 推理树如何变成稠密技能更新证据,降低 token 成本 | 精读 |
| 4 | OpenLoopEvolve: A Verifiable Self-Evolution Framework for Loop Policies in Long-Horizon Complex Tasks | 在线/离线自进化如何做门控、回滚和版本发布 | 精读 |
| 5 | SHE: Trajectory-driven Safety Harness Evolution for LLM Agents | 安全 harness 如何按轨迹归因局部进化 | 精读 |
| 6 | Controlled Memory Interference in Continual LLM Agents | 记忆的干扰、更新权威性与投毒边界 | 精读 |
| 7 | Not Worth Another Token: Marginal Value Estimation for Efficient Deep Research Agents | 哪一阶段剪枝最省 token,又不把验证当成剪枝 | 精读 |
| 8 | Agentic Auto-Research is Fuzz Testing | 研究代理为何应把中间进展信号当作 fuzzing 覆盖率 | 邻近跟踪 |
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.