<?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>推理机制 on MessageDaily</title><link>https://inkeast.github.io/MessageDaily/tags/%E6%8E%A8%E7%90%86%E6%9C%BA%E5%88%B6/</link><description>Recent content in 推理机制 on MessageDaily</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Sun, 06 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://inkeast.github.io/MessageDaily/tags/%E6%8E%A8%E7%90%86%E6%9C%BA%E5%88%B6/index.xml" rel="self" type="application/rss+xml"/><item><title>Locked at the Entrance 精读：RLVR 的多样性坍缩发生在推理的门口</title><link>https://inkeast.github.io/MessageDaily/posts/2026-09-06-locked-at-the-entrance-rlvr-paper-reading/</link><pubDate>Sun, 06 Sep 2026 00:00:00 +0000</pubDate><guid>https://inkeast.github.io/MessageDaily/posts/2026-09-06-locked-at-the-entrance-rlvr-paper-reading/</guid><description>RLVR 提升 pass@1 却坍缩解空间已是共识，但坍缩发生在哪一步始终未知。上海大学×伯明翰大学在 Countdown 任务上穷举解空间、按首操作数+算子划分入口族，把求解分解为 access×execution：PPO 覆盖 0.337→0.111，首算术操作前的似然偏移是下游的 11–16 倍，塞一个入口前缀就能让低覆盖族完成率 0.018→0.212——能力还在，只是不再进门。后层参数插值恢复 37% 覆盖且 pass@1 零损失。</description></item></channel></rss>