<?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>Visual-Reasoning on Xu'Blog</title><link>https://xuquant.com/tags/visual-reasoning/</link><description>Recent content in Visual-Reasoning on Xu'Blog</description><image><title>Xu'Blog</title><url>https://xuquant.com/images/profile.jpg</url><link>https://xuquant.com/images/profile.jpg</link></image><generator>Hugo -- 0.152.2</generator><language>en</language><lastBuildDate>Thu, 30 Apr 2026 20:00:00 +0800</lastBuildDate><atom:link href="https://xuquant.com/tags/visual-reasoning/index.xml" rel="self" type="application/rss+xml"/><item><title>DeepSeek 以视觉原语思考：让多模态大模型学会「用手指着推理」</title><link>https://xuquant.com/posts/deepseek-thinking-with-visual-primitives/</link><pubDate>Thu, 30 Apr 2026 20:00:00 +0800</pubDate><guid>https://xuquant.com/posts/deepseek-thinking-with-visual-primitives/</guid><description>深度解读 DeepSeek 联合北大/清华提出的「以视觉原语思考」新范式：用坐标和边界框替代自然语言描述，解决多模态模型推理过程中的指代歧义问题，迷宫导航任务领先 GPT-5.4 达 17 个百分点。</description></item></channel></rss>