Papers2

#spatiotemporal reasoning

Learning to Reason in 4D: Dynamic Spatial Understanding for Vision Language Models

Shengchao Zhou, Yuxin Chen et al.Dec 23arXiv

The paper tackles a big blind spot in vision-language models: understanding how objects move and relate in 3D over time (dynamic spatial reasoning, or DSR).

#dynamic spatial reasoning#vision-language models#4D understanding

4D-RGPT: Toward Region-level 4D Understanding via Perceptual Distillation

Intermediate

Chiao-An Yang, Ryo Hachiuma et al.Dec 18arXiv

This paper teaches a video-understanding AI to think in 3D plus time (4D) so it can answer questions about specific objects moving in videos.

#4D perception#multimodal large language models#perceptual distillation