<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
  <channel>
    <title>(&#39;&#39;, &#39;cs.CLcs.AI&#39;) - Tag - leoleils-blog</title>
    <link>https://leoleils.top/tags/cs.clcs.ai/</link>
    <description>(&#39;&#39;, &#39;cs.CLcs.AI&#39;) - Tag - leoleils-blog</description>
    <generator>Hugo -- gohugo.io</generator><language>zh-cn</language><lastBuildDate>Wed, 06 May 2026 12:00:00 &#43;0800</lastBuildDate><atom:link href="https://leoleils.top/tags/cs.clcs.ai/" rel="self" type="application/rss+xml" /><item>
  <title>📄 Safety and accuracy follow different scaling laws in clinica</title>
  <link>https://leoleils.top/posts/ai-paper/paper-2026-05-06-safety-and-accuracy-follow-different-scaling-laws-/</link>
  <pubDate>Wed, 06 May 2026 12:00:00 &#43;0800</pubDate>
  <author>Author</author>
  <guid>https://leoleils.top/posts/ai-paper/paper-2026-05-06-safety-and-accuracy-follow-different-scaling-laws-/</guid>
  <description><![CDATA[📖 论文基本信息 项目 内容 标题 Safety and accuracy follow different scaling laws in clinical large language models 作者 Sebastian Wind, Tri-Thien Nguyen, Jeta Sopa, Mahshad Lotfinia, Sebastian Bickelhaup&hellip; 机构 待补充 发布时间 2026-05-05 arXiv 编号 2605.04039v1 PDF 下载 点击下载 💡 一句话总结 Clinical LLMs are often scaled by increasing model size, context length, retrieval complexity, or inference-time compute, with the implicit expectation that higher accuracy implies safer behavior. Thi&hellip;
🎯 研究背景 这部分需要深入分析论文要解决的核心问题]]></description>
</item>
<item>
  <title>📄 Turning the TIDE: Cross-Architecture Distillation for Diffus</title>
  <link>https://leoleils.top/posts/ai-paper/paper-2026-04-30-turning-the-tide-cross-architecture-distillation-f/</link>
  <pubDate>Thu, 30 Apr 2026 12:00:00 &#43;0800</pubDate>
  <author>Author</author>
  <guid>https://leoleils.top/posts/ai-paper/paper-2026-04-30-turning-the-tide-cross-architecture-distillation-f/</guid>
  <description><![CDATA[📖 论文基本信息 项目 内容 标题 Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models 作者 Gongbo Zhang, Wen Wang, Ye Tian, Li Yuan 机构 待补充 发布时间 2026-04-29 arXiv 编号 2604.26951v1 PDF 下载 点击下载 💡 一句话总结 Diffusion large language models (dLLMs) offer parallel decoding and bidirectional context, but state-of-the-art dLLMs require billions of parameters for competitive performance. While existing distill&hellip;
🎯 研究背景 这部分需要深入分析论文要解决的核心问题
现有挑战 待补充：现有方法的局限性 研究动机 待补充：为什么这个问题重要 🔬 核心创新 这部分需要提炼 3-5 个技术突破点]]></description>
</item>
<item>
  <title>📄 SpeechParaling-Bench: A Comprehensive Benchmark for Paraling</title>
  <link>https://leoleils.top/posts/ai-paper/paper-2026-04-23-speechparaling-bench-a-comprehensive-benchmark-for/</link>
  <pubDate>Thu, 23 Apr 2026 12:00:00 &#43;0800</pubDate>
  <author>Author</author>
  <guid>https://leoleils.top/posts/ai-paper/paper-2026-04-23-speechparaling-bench-a-comprehensive-benchmark-for/</guid>
  <description><![CDATA[📖 论文基本信息 项目 内容 标题 SpeechParaling-Bench: A Comprehensive Benchmark for Paralinguistic-Aware Speech Generation 作者 Ruohan Liu, Shukang Yin, Tao Wang, Dong Zhang, Weiji Zhuang&hellip; 机构 待补充 发布时间 2026-04-22 arXiv 编号 2604.20842v1 PDF 下载 点击下载 💡 一句话总结 Paralinguistic cues are essential for natural human-computer interaction, yet their evaluation in Large Audio-Language Models (LALMs) remains limited by coarse feature coverage and the inherent subjec&hellip;
🎯 研究背景 这部分需要深入分析论文要解决的核心问题
现有挑战 待补充：现有方法的局限性 研究动机 待补充：为什么这个问题重要 🔬 核心创新 这部分需要提炼 3-5 个技术突破点]]></description>
</item>
<item>
  <title>📄 Early Stopping for Large Reasoning Models via Confidence Dyn</title>
  <link>https://leoleils.top/posts/ai-paper/paper-2026-04-07-early-stopping-for-large-reasoning-models-via-conf/</link>
  <pubDate>Tue, 07 Apr 2026 12:00:00 &#43;0800</pubDate>
  <author>Author</author>
  <guid>https://leoleils.top/posts/ai-paper/paper-2026-04-07-early-stopping-for-large-reasoning-models-via-conf/</guid>
  <description><![CDATA[📖 论文基本信息 项目 内容 标题 Early Stopping for Large Reasoning Models via Confidence Dynamics 作者 Parsa Hosseini, Sumit Nawathe, Mahdi Salmani, Meisam Razaviyayn, Soheil Feizi 机构 待补充 发布时间 2026-04-06 arXiv 编号 2604.04930v1 PDF 下载 点击下载 💡 一句话总结 Large reasoning models rely on long chain-of-thought generation to solve complex problems, but extended reasoning often incurs substantial computational cost and can even degrade performance due to ov&hellip;
🎯 研究背景 这部分需要深入分析论文要解决的核心问题
现有挑战 待补充：现有方法的局限性 研究动机 待补充：为什么这个问题重要 🔬 核心创新 这部分需要提炼 3-5 个技术突破点]]></description>
</item>
</channel>
</rss>
