Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning

Summary

Multimodal large language models increasingly use visual chain-of-thought (Visual CoT), which generates intermediate reasoning images for spatial, temporal, and embodied reasoning, but introduces significant inference overhead. This paper introduces Internalized Visual Thinking (IVT), a post-training framework that jointly optimizes textual prediction and next-embedding prediction on unlabeled videos, enabling the model to directly generate answers at inference without synthesizing future frames. Across all six evaluation settings, IVT improves over direct-answer fine-tuning, and compared with explicit Visual CoT, it achieves comparable or better performance while reducing average end-to-end latency by more than 5x. The findings suggest that explicit pixel-space generation at inference may not be necessary for effective proactive video reasoning.