<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>PaperWeekly on 文字轨迹</title><link>https://ixxmu.github.io/tags/paperweekly/</link><description>Recent content in PaperWeekly on 文字轨迹</description><generator>Hugo</generator><language>zh-cn</language><lastBuildDate>Sat, 20 Dec 2025 16:16:40 +0000</lastBuildDate><atom:link href="https://ixxmu.github.io/tags/paperweekly/index.xml" rel="self" type="application/rss+xml"/><item><title>OpenAI官方论文“泄密”GPT-5：RL到底有没有教坏CoT？万字深度实测</title><link>https://ixxmu.github.io/posts/2025-12/openai%E5%AE%98%E6%96%B9%E8%AE%BA%E6%96%87_%E6%B3%84%E5%AF%86_gpt_5_rl%E5%88%B0%E5%BA%95%E6%9C%89%E6%B2%A1%E6%9C%89%E6%95%99%E5%9D%8Fcot_%E4%B8%87%E5%AD%97%E6%B7%B1%E5%BA%A6%E5%AE%9E%E6%B5%8B/</link><pubDate>Sat, 20 Dec 2025 16:16:40 +0000</pubDate><guid>https://ixxmu.github.io/posts/2025-12/openai%E5%AE%98%E6%96%B9%E8%AE%BA%E6%96%87_%E6%B3%84%E5%AF%86_gpt_5_rl%E5%88%B0%E5%BA%95%E6%9C%89%E6%B2%A1%E6%9C%89%E6%95%99%E5%9D%8Fcot_%E4%B8%87%E5%AD%97%E6%B7%B1%E5%BA%A6%E5%AE%9E%E6%B5%8B/</guid><description>OpenAI官方论文“泄密”GPT-5：RL到底有没有教坏CoT？万字深度实测 by PaperWeekly 你的模型是在真思考，还是为了讨好 Reward Model 在演戏？</description></item><item><title>Google Scholar被玩坏：10篇“水文”刷出600+引用，H-index还能信吗？</title><link>https://ixxmu.github.io/posts/2025-12/google_scholar%E8%A2%AB%E7%8E%A9%E5%9D%8F_10%E7%AF%87_%E6%B0%B4%E6%96%87_%E5%88%B7%E5%87%BA600_%E5%BC%95%E7%94%A8_h_index%E8%BF%98%E8%83%BD%E4%BF%A1%E5%90%97_/</link><pubDate>Thu, 11 Dec 2025 23:38:59 +0000</pubDate><guid>https://ixxmu.github.io/posts/2025-12/google_scholar%E8%A2%AB%E7%8E%A9%E5%9D%8F_10%E7%AF%87_%E6%B0%B4%E6%96%87_%E5%88%B7%E5%87%BA600_%E5%BC%95%E7%94%A8_h_index%E8%BF%98%E8%83%BD%E4%BF%A1%E5%90%97_/</guid><description>Google Scholar被玩坏：10篇“水文”刷出600+引用，H-index还能信吗？ by PaperWeekly 现在的 Google Scholar，漏洞大得像个筛子。</description></item></channel></rss>