New PhyCheck dataset aims to improve Video-LLM understanding of physical laws

Researchers have introduced PhyCheck, a new dataset designed to improve the physical law understanding of Video Large Language Models (VideoLLMs). The dataset includes coarse-grained questions about physical compliance and fine-grained questions about specific physical details. Experiments fine-tuning Qwen2.5-VL with PhyCheck data showed significant improvements in physical consistency understanding, though models still struggle with incorporating additional causal conditions. AI

IMPACT This dataset could lead to more robust Video-LLMs capable of understanding physical interactions, crucial for embodied AI and world models.

RANK_REASON The cluster describes a new academic dataset and associated research paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• PhyCheck
• Qwen2.5-VL
• Video LLMs

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →