Paper 2510.23473
Video-Thinker: Sparking "Thinking with Videos" via Reinforcement Learning
- Published
- Oct 2025
- Research lab
- Independent
- Citations
- 21
- GitHub
- 160 stars
01 In brief
Summary
Video-Thinker is a framework that extends the 'Thinking with Images' paradigm to video reasoning by enabling Multimodal Large Language Models (MLLMs) to autonomously use intrinsic 'grounding' (temporal localization) and 'captioning' (visual comprehension) capabilities within chain-of-thought reasoning, eliminating the need for external tools.
The authors constructed a 10K-sample dataset (Video-Thinker-10K) via a hindsight-curation pipeline, then trained with Supervised Fine-Tuning (SFT) followed by Group Relative Policy Optimization (GRPO).
Video-Thinker-7B achieved state-of-the-art performance among 7B-sized MLLMs on out-of-domain benchmarks (Video-Holmes: 43.22%, CG-Bench-Reasoning: 33.25%, VRBench: 80.69%) and in-domain tasks, outperforming baselines like Video-R1-7B.
Ablations showed GRPO significantly improves generalization over SFT alone, and the model excels in grounding (mIoU 48.22%) and captioning (average 17.11%) tasks.
The work demonstrates that effective video reasoning can be achieved with only 10K curated samples, contrasting with prior methods requiring over 160K samples.
02 From the paper
Abstract
Recent advances in image reasoning methods, particularly "Thinking with Images", have demonstrated remarkable success in Multimodal Large Language Models (MLLMs); however, this dynamic reasoning paradigm has not yet been extended to video reasoning tasks. In this paper, we propose Video-Thinker, which empowers MLLMs to think with videos by autonomously leveraging their intrinsic "grounding" and "captioning" capabilities to generate reasoning clues throughout the inference process. To spark this capability, we construct Video-Thinker-10K, a curated dataset featuring autonomous tool usage within chain-of-thought reasoning sequences. Our training strategy begins with Supervised Fine-Tuning (SFT) to learn the reasoning format, followed by Group Relative Policy Optimization (GRPO) to strengthen this reasoning capability. Through this approach, Video-Thinker enables MLLMs to autonomously navigate grounding and captioning tasks for video reasoning, eliminating the need for constructing and calling external tools. Extensive experiments demonstrate that Video-Thinker achieves significant performance gains on both in-domain tasks and challenging out-of-domain video reasoning benchmarks, including Video-Holmes, CG-Bench-Reasoning, and VRBench. Our Video-Thinker-7B substantially outperforms existing baselines such as Video-R1 and establishes state-of-the-art performance among 7B-sized MLLMs.