https://spidermonk7.github.io/tags/multimodal-ai/Multimodal AIHugo Blox Builder (https://hugoblox.com)en-usFri, 03 Apr 2026 00:00:00 +0000https://spidermonk7.github.io/media/icon_hu7729264130191091259.pnghttps://spidermonk7.github.io/tags/multimodal-ai/https://spidermonk7.github.io/ongoing-projects/vidnum-1-4k/Fri, 03 Apr 2026 00:00:00 +0000https://spidermonk7.github.io/ongoing-projects/vidnum-1-4k/<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.</p> <!-- Official page: <https://vidnumteam.github.io> -->