VideoRFT: Incentivizing Video Reasoning Capability in MLLMs via Reinforced Fine-Tuning

TL;DR

VideoRFT introduces a novel reinforcement fine-tuning approach to enhance multi-modal large language models' video reasoning abilities, leveraging new datasets and a semantic consistency reward to achieve state-of-the-art results.

Contribution

The paper presents VideoRFT, a new method extending reinforcement fine-tuning to video reasoning in MLLMs, including a reasoning dataset and a semantic consistency reward.

Findings

Achieved state-of-the-art performance on six video reasoning benchmarks.

Developed two new datasets: VideoRFT-CoT-102K and VideoRFT-RL-310K.

Introduced a semantic consistency reward to improve reasoning grounded in visual evidence.

Abstract

Reinforcement fine-tuning (RFT) has shown great promise in achieving humanlevel reasoning capabilities of Large Language Models (LLMs), and has recently been extended to MLLMs. Nevertheless, reasoning about videos, which is a fundamental aspect of human intelligence, remains a persistent challenge due to the complex logic, temporal and causal structures inherent in video data. To fill this gap, we propose VideoRFT, a novel approach that extends the RFT paradigm to cultivate human-like video reasoning capabilities in MLLMs. VideoRFT follows the standard two-stage scheme in RFT: supervised fine-tuning (SFT) with chain-of-thought (CoT) annotations, followed by reinforcement learning (RL) to improve generalization. A central challenge to achieve this in the video domain lies in the scarcity of large-scale, high-quality video CoT datasets. We address this by building a multi-expert-driven, cognition-inspired CoT curation pipeline. First, we devise a cognition-inspired prompting strategy to elicit a reasoning LLM to generate preliminary CoTs based solely on rich, structured, and literal representations of video content. Subsequently, these CoTs are revised by a MLLM conditioned on the actual video, ensuring visual consistency and reducing visual hallucinations. This pipeline results in two new datasets, i.e.VideoRFT-CoT-102K for SFT and VideoRFT-RL-310K for RL. To further strengthen the RL phase, we introduce a novel semantic-consistency reward that explicitly promotes the alignment between textual reasoning and visual evidence. This reward encourages the model to produce coherent, context-aware reasoning outputs grounded in visual input. Extensive experiments show that VideoRFT achieves state-of-the-art performance on six video reasoning benchmarks.