Paper proposes answer self-consistency method for CVPR 2026 VidLLMs Challenge

Arxiv CS.CV (computer vision)

The method uses margin-triggered question re-arbitration to improve visual relational reasoning in videos. Submitted to Track 2 of the CVPR 2026 VidLLMs Challenge.

llm

Arxiv CS.CV (computer vision)·Tomoya MiyazawaHiroyasu Okuno

7 days ago