VideoThinker is a causal-inspired framework that enables lightweight multimodal language models (3B parameters) to achieve robust video reasoning. It addresses the phenomenon of "perceptual bias," where reinforcement learning can compel lightweight models to adopt perceptual shortcuts from data rather than developing genuine reasoning abilities.
VideoThinker-R1 establishes a new state-of-the-art in video reasoning efficiency. Using only 1K training samples and no Supervised Fine-Tuning (SFT), it:
1@inproceedings{wu2026videothinker,
2 title={Beyond Perceptual Shortcuts: Causal-Inspired Debiasing Optimization for Generalizable Video Reasoning in Lightweight MLLMs},
3 author={Wu, Jingze and Zhang, Quan and Suo, Hongfei and Cai, Zeqiang and Chen, Hongbo},
4 booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
5 year={2026}
6}
