TriPhysBench evaluates physical intelligence coordination in vision-language models
Original title: TriPhysBench: Evaluating Physical Intelligence for Embodied Agents
Benchmark shows top vision-language models achieve only 50.38% accuracy on integrated physical tasks.
Embodied agents require coordinated physical perception, knowledge retrieval, and reasoning. TriPhysBench provides 410 questions from 200 videos to test these capabilities together. Existing models perform poorly on the joint tasks. TriPhys learns a unified policy that improves answer accuracy by 17.7% relative over the Qwen3.6-27B backbone.
- TriPhysBench contains 410 questions from 200 videos
- Best model accuracy is 50.38%
- TriPhys uses intermediate reasoning and tool feedback
- Improves scores across three Qwen configurations
From the original article
1 min read at arxiv.orgEmbodied agents need physical intelligence to understand their surroundings through physical perception, physical knowledge retrieval, and physical reasoning. These capabilities must work together to interpret events and make quantitative judgments.
By Zhongxiang Sun · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.