SpaceR: Reinforcing MLLMs in Video Spatial Reasoning

Kun Ouyang, Yuanxin Liu, Haoning Wu, Yi Liu, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun

Keywords: Reinforcement Learning, Video Spatial Reasoning, Multimodal Large Language Models

TL;DR: We introduce a high-quality dataset SpaceR-151k and propose SpaceR trained via our SG-RLVR framework towards video spatial reasoning.

Abstract:

Video spatial reasoning, which involves inferring the underlying spatial structure from observed video frames, poses a significant challenge for existing Multimodal Large Language Models (MLLMs). This limitation stems primarily from 1) the absence of high-quality datasets for this task, and 2) the lack of effective training strategies to develop spatial reasoning capabilities. Motivated by the success of Reinforcement Learning with Verifiable Reward (RLVR) in unlocking LLM reasoning abilities, this work aims to improve MLLMs in video spatial reasoning through the RLVR paradigm. To this end, we introduce the **SpaceR** framework. First, we present **SpaceR-151k**, a dataset with 91k questions spanning diverse spatial reasoning scenarios with verifiable answers, and 60k samples for maintaining general multimodal understanding. Second, we propose **Spatially-Guided RLVR (SG-RLVR)**, a novel reinforcement learning approach that extends Group Relative Policy Optimization (GRPO) with a novel map imagination mechanism, which encourages the model to infer spatial layouts in the thinking process, thereby facilitating more effective spatial reasoning.
Extensive experiments demonstrate that SpaceR achieves state-of-the-art performance on spatial reasoning benchmarks (e.g., VSI-Bench, STI-Bench, and SPAR-Bench), while showing competitive results on video understanding benchmarks (e.g., Video-MME, TempCompass, and LongVideoBench).
Remarkably, SpaceR surpasses the advanced GPT-4o by 11.6\% accuracy on VSI-Bench and is on par with the leading proprietary model Gemini-2.0-Flash, highlighting the effectiveness of our SpaceR-151k dataset and SG-RLVR in reinforcing spatial reasoning ability of MLLMs.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 12588