<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>多模态 on AI 早报</title><link>https://ai-news.example.com/tags/%E5%A4%9A%E6%A8%A1%E6%80%81/</link><description>Recent content in 多模态 on AI 早报</description><generator>Hugo</generator><language>zh-CN</language><lastBuildDate>Sat, 24 Oct 2026 00:29:15 +0800</lastBuildDate><atom:link href="https://ai-news.example.com/tags/%E5%A4%9A%E6%A8%A1%E6%80%81/feed.xml" rel="self" type="application/rss+xml"/><item><title>梁文锋押注具身智能，DeepSeek招募Swin Transformer作者布局具身大脑</title><link>https://ai-news.example.com/articles/deepseekswin-transformer/</link><pubDate>Sat, 24 Oct 2026 00:29:15 +0800</pubDate><guid>https://ai-news.example.com/articles/deepseekswin-transformer/</guid><description>DeepSeek创始人梁文锋投资并招募Swin Transformer核心作者，正式启动具身智能大脑方向布局，标志着这家以大模型著称的公司在多模态与机器人领域迈出关键一步。</description></item><item><title>iPhone断网跑35B多模态！国产开源新模型仅占3G内存</title><link>https://ai-news.example.com/articles/iphone35b3g/</link><pubDate>Sat, 17 Oct 2026 00:18:36 +0800</pubDate><guid>https://ai-news.example.com/articles/iphone35b3g/</guid><description>国产开源模型实现重大突破，可在iPhone断网状态下运行35B参数的多模态大模型，内存占用仅3G，大幅降低端侧AI部署门槛。</description></item><item><title>何恺明团队发布多模态Harness框架</title><link>https://ai-news.example.com/articles/harness/</link><pubDate>Mon, 12 Oct 2026 01:12:20 +0800</pubDate><guid>https://ai-news.example.com/articles/harness/</guid><description>何恺明团队推出多模态Harness框架，旨在统一和提升多模态AI模型的训练与评估效率，为通用人工智能研究提供新工具。</description></item></channel></rss>