Preprint
Sep 2026
Aligning Thoughts with Answers: Probability Rewards to Tame Thinking Drift
This paper studies \textbf{thinking--answer consistency} in vision-language models. We focus on Visual Intention Grounding, where a model infers a target object based on a human intention query and predicts a bounding box. We reveal that previous IoU-based reinforcement learning (RL) frameworks suffer from ``thinking d...
Peng-Zhan Sun, Shiu-hong Kao, Shi-Jie Li et al.
· 1 citation