VideoChat-R1：通過強化微調提升時空感知能力

摘要

近期強化學習的進展顯著提升了多模態大型語言模型（MLLMs）的推理能力。雖然群組相對策略優化（GRPO）和基於規則的獎勵機制在文本和圖像領域展現出潛力，但它們在視頻理解中的應用仍有限。本文系統地探討了將GRPO應用於視頻MLLMs的強化微調（RFT），旨在增強時空感知能力，同時保持模型的通用能力。我們的實驗表明，RFT在特定任務改進上具有極高的數據效率。通過在有限樣本上進行時空感知目標的多任務RFT，我們開發了VideoChat-R1，這是一個強大的視頻MLLM，在時空感知任務上達到了最先進的性能，且未犧牲聊天能力，同時展現出新興的時空推理能力。與Qwen2.5-VL-7B相比，VideoChat-R1在時間定位（+31.8）和目標跟踪（+31.2）等任務中的性能提升了數倍。此外，它在通用問答基準測試如VideoMME（+0.9）、MVBench（+1.0）和Perception Test（+0.9）上也有顯著提升。我們的研究結果強調了RFT在視頻MLLMs專項任務增強中的潛力。我們希望這項工作能為未來視頻MLLMs的強化學習研究提供寶貴的見解。

English

Recent advancements in reinforcement learning have significantly advanced the reasoning capabilities of multimodal large language models (MLLMs). While approaches such as Group Relative Policy Optimization (GRPO) and rule-based reward mechanisms demonstrate promise in text and image domains, their application to video understanding remains limited. This paper presents a systematic exploration of Reinforcement Fine-Tuning (RFT) with GRPO for video MLLMs, aiming to enhance spatio-temporal perception while maintaining general capabilities. Our experiments reveal that RFT is highly data-efficient for task-specific improvements. Through multi-task RFT on spatio-temporal perception objectives with limited samples, we develop VideoChat-R1, a powerful video MLLM that achieves state-of-the-art performance on spatio-temporal perception tasks without sacrificing chat ability, while exhibiting emerging spatio-temporal reasoning abilities. Compared to Qwen2.5-VL-7B, VideoChat-R1 boosts performance several-fold in tasks like temporal grounding (+31.8) and object tracking (+31.2). Additionally, it significantly improves on general QA benchmarks such as VideoMME (+0.9), MVBench (+1.0), and Perception Test (+0.9). Our findings underscore the potential of RFT for specialized task enhancement of Video MLLMs. We hope our work offers valuable insights for future RL research in video MLLMs.

VideoChat-R1：通過強化微調提升時空感知能力

VideoChat-R1: Enhancing Spatio-Temporal Perception via Reinforcement Fine-Tuning

摘要

Support