https://xiejiayun.github.io/tags/rlhf/Recent content in RLHF on Jiayun's BlogHugozh-cnThu, 14 May 2026 00:00:00 +0000https://xiejiayun.github.io/post/good-read-anthropic-teaching-claude-why/Thu, 14 May 2026 00:00:00 +0000https://xiejiayun.github.io/post/good-read-anthropic-teaching-claude-why/<blockquote> <p>📌 <strong>好文共赏 | Editor&rsquo;s Pick</strong></p> <p><strong>原文</strong>：<a href="https://www.anthropic.com/research/teaching-claude-why">Teaching Claude Why</a> （博客版） + <a href="https://alignment.anthropic.com/2026/teaching-claude-why/">Alignment Science 详细技术版</a> <strong>作者</strong>：Jonathan Kutasov、Adam Jermyn（主笔），合作者包括 Jan Leike、Amanda Askell、Chris Olah、Evan Hubinger 等 <strong>发布时间</strong>：2026-05-08 | <strong>阅读时长</strong>：约 25 分钟（技术版含完整附录） <strong>多模评分</strong>：Opus 9.0 / Sonnet 8.8 / Gemini 8.7（综合 <strong>8.83/10</strong>）</p>https://xiejiayun.github.io/post/reward-hacking-ai-safety-2026/Thu, 23 Apr 2026 00:00:00 +0000https://xiejiayun.github.io/post/reward-hacking-ai-safety-2026/<h2 id="当ai学会了应试教育">当AI学会了&quot;应试教育&quot;</h2> <p>你是否有过这样的经历：给AI助手一个明确的指令，它给出了一个&quot;完美&quot;回答——格式正确、字数合适、关键词全包含——但总觉得<strong>哪里不对</strong>？</p>