<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>探针 on MessageDaily</title><link>https://inkeast.github.io/MessageDaily/tags/%E6%8E%A2%E9%92%88/</link><description>Recent content in 探针 on MessageDaily</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Sun, 21 Jun 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://inkeast.github.io/MessageDaily/tags/%E6%8E%A2%E9%92%88/index.xml" rel="self" type="application/rss+xml"/><item><title>Probe-and-Refine Tuning of Repository Guidance for Coding Agents 精读</title><link>https://inkeast.github.io/MessageDaily/posts/2026-06-21-probe-and-refine-tuning-%E8%AE%BA%E6%96%87%E7%B2%BE%E8%AF%BB/</link><pubDate>Sun, 21 Jun 2026 00:00:00 +0000</pubDate><guid>https://inkeast.github.io/MessageDaily/posts/2026-06-21-probe-and-refine-tuning-%E8%AE%BA%E6%96%87%E7%B2%BE%E8%AF%BB/</guid><description>深度精读 Williams College 的 Probe-and-Refine Tuning 论文——一种用合成 bug 修复探针迭代打磨仓库引导文件（AGENTS.md）的极简方法。无需智能体循环、无需工具调用、无需强化学习，仅靠约 22 次单轮 LLM 调用就让 SWE-bench Verified 解决率从 25.5% 提升到 33.0%。核心发现是：改进的不是补丁质量而是覆盖率，指导文件让 Agent 跑到了正确的文件。</description></item></channel></rss>