M2-Reasoning: Empowering MLLMs with Unified General and Spatial Reasoning
This addresses a critical gap in MLLMs for real-world applications requiring spatial reasoning, representing a strong specific gain rather than a foundational breakthrough.
The paper tackles the problem of multimodal large language models struggling with dynamic spatial interactions by introducing M2-Reasoning-7B, which achieves state-of-the-art performance across 8 benchmarks through a novel data pipeline generating 294.2K samples and a dynamic multi-task training strategy.
Recent advancements in Multimodal Large Language Models (MLLMs), particularly through Reinforcement Learning with Verifiable Rewards (RLVR), have significantly enhanced their reasoning abilities. However, a critical gap persists: these models struggle with dynamic spatial interactions, a capability essential for real-world applications. To bridge this gap, we introduce M2-Reasoning-7B, a model designed to excel in both general and spatial reasoning. Our approach integrates two key innovations: (1) a novel data pipeline that generates 294.2K high-quality data samples (168K for cold-start fine-tuning and 126.2K for RLVR), which feature logically coherent reasoning trajectories and have undergone comprehensive assessment; and (2) a dynamic multi-task training strategy with step-wise optimization to mitigate conflicts between data, and task-specific rewards for delivering tailored incentive signals. This combination of curated data and advanced training allows M2-Reasoning-7B to set a new state-of-the-art (SOTA) across 8 benchmarks, showcasing superior performance in both general and spatial reasoning domains.