<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Multimodal AI | Shaoyang Cui</title><link>https://spidermonk7.github.io/tags/multimodal-ai/</link><atom:link href="https://spidermonk7.github.io/tags/multimodal-ai/index.xml" rel="self" type="application/rss+xml"/><description>Multimodal AI</description><generator>Hugo Blox Builder (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Fri, 03 Apr 2026 00:00:00 +0000</lastBuildDate><image><url>https://spidermonk7.github.io/media/icon_hu7729264130191091259.png</url><title>Multimodal AI</title><link>https://spidermonk7.github.io/tags/multimodal-ai/</link></image><item><title>VidNum1.4K - A Comprehensive Benchmark for Video-based Numerical Reasoning</title><link>https://spidermonk7.github.io/ongoing-projects/vidnum-1-4k/</link><pubDate>Fri, 03 Apr 2026 00:00:00 +0000</pubDate><guid>https://spidermonk7.github.io/ongoing-projects/vidnum-1-4k/</guid><description>&lt;p>This research introduces VNum, a comprehensive VideoQA benchmark containing 1,379 human-annotated video-question pairs designed to test multi-step numerical reasoning in Vision-Language Models (VLMs). Moving beyond simple counting, VNum spans diverse real-world environments to quantify objects, actions, and events through a unique three-level hierarchy.&lt;/p>
&lt;!-- Official page: &lt;https://vidnumteam.github.io> --></description></item></channel></rss>