<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>AI HOT 日报</title>
    <link>https://aihot-cf.pages.dev/feed/daily.xml</link>
    <description>每日 AI 领域重要动态</description>
    <language>zh-CN</language>
    <atom:link href="https://aihot-cf.pages.dev/feed/daily.xml" rel="self" type="application/rss+xml"/>
    <lastBuildDate>2026-05-28T19:19:16.050Z</lastBuildDate>
    <generator>AIHOT on Cloudflare</generator>
    <item>
      <title>VibeSearchBench：面向真实世界中长期主动搜索的评测基准</title>
      <link>https://arxiv.org/abs/2605.27882</link>
      <guid isPermaLink="true">https://arxiv.org/abs/2605.27882</guid>
      <pubDate>2026-05-27T22:08:22.000Z</pubDate>
      <description>研究提出了&quot;VibeSearch&quot;范式并发布了VibeSearchBench，包含200个手工策划的双语任务。对七个前沿模型的测试显示，所有模型在VibeSearch任务上表现均不充分（最佳F1分数为30.30）。</description>
    </item>
    <item>
      <title>英伟达推出 AI 框架 Polar，让 Codex 跑分暴涨 594.74%</title>
      <link>https://www.ithome.com/0/956/293.htm</link>
      <guid isPermaLink="true">https://www.ithome.com/0/956/293.htm</guid>
      <pubDate>2026-05-27T14:08:22.000Z</pubDate>
      <description>英伟达研究团队开源了智能体强化学习框架 Polar。基于 Qwen3.5-4B 模型，将 Codex 在 SWE-Bench Verified 上的 pass@1 分数从 3.8% 提升至 26.4%。</description>
    </item>
    <item>
      <title>OpenRouter 获得1.13亿美元B轮融资</title>
      <link>https://openrouter.ai/announcements/series-b</link>
      <guid isPermaLink="true">https://openrouter.ai/announcements/series-b</guid>
      <pubDate>2026-05-27T12:08:22.000Z</pubDate>
      <description>AI模型聚合平台OpenRouter宣布完成1.13亿美元B轮融资。本轮融资由CapitalG领投，NVentures、ServiceNow Ventures等多家机构参投。</description>
    </item>
    <item>
      <title>DenoiseRL：通过恢复嘈杂前缀来引导推理模型</title>
      <link>https://arxiv.org/abs/2605.28421</link>
      <guid isPermaLink="true">https://arxiv.org/abs/2605.28421</guid>
      <pubDate>2026-05-27T10:08:22.000Z</pubDate>
      <description>DenoiseRL是一种强化学习框架，旨在提升大语言模型的推理能力。它无需依赖更强的教师模型，而是通过在弱模型产生的失败推理轨迹上进行基于恢复的优化来直接学习。</description>
    </item>
    <item>
      <title>pgvector驱动的语义、混合、稀疏与量化向量搜索系统构建编码指南</title>
      <link>https://www.marktechpost.com/2026/05/28/a-coding-guide-to-implement-a-pgvector-powered-semantic-hybrid-sparse-and-quantized-vector-search-system</link>
      <guid isPermaLink="true">https://www.marktechpost.com/2026/05/28/a-coding-guide-to-implement-a-pgvector-powered-semantic-hybrid-sparse-and-quantized-vector-search-system</guid>
      <pubDate>2026-05-27T08:08:22.000Z</pubDate>
      <description>本教程在Google Colab中构建一个完整的pgvector实验环境，展示PostgreSQL如何作为向量数据库服务于现代AI应用。</description>
    </item>
  </channel>
</rss>