V-tableR1: Process-Supervised Multimodal Table Reasoning with Critic-Guided Policy Optimization
This addresses the issue of unreliable reasoning in multimodal AI for tasks like table analysis, though it is incremental as it focuses on a specific visual domain.
The paper tackles the problem of multimodal large language models relying on superficial pattern matching for visual reasoning by introducing V-tableR1, a process-supervised reinforcement learning framework that uses critic-guided feedback to enforce verifiable reasoning; it achieves state-of-the-art accuracy on complex tabular benchmarks, outperforming models up to 18x its size.
We introduce V-tableR1, a process-supervised reinforcement learning framework that elicits rigorous, verifiable reasoning from multimodal large language models (MLLMs). Current MLLMs trained solely on final outcomes often treat visual reasoning as a black box, relying on superficial pattern matching rather than performing rigorous multi-step inference. While Reinforcement Learning with Verifiable Rewards could enforce transparent reasoning trajectories, extending it to visual domains remains severely hindered by the ambiguity of grounding abstract logic into continuous pixel space. We solve this by leveraging the deterministic grid structure of tables as an ideal visual testbed. V-tableR1 employs a specialized critic VLM to provide dense, step-level feedback on the explicit visual chain-of-thought generated by a policy VLM. To optimize this system, we propose Process-Guided Direct Alignment Policy Optimization (PGPO), a novel RL algorithm integrating process rewards, decoupled policy constraints, and length-aware dynamic sampling. Extensive evaluations demonstrate that V-tableR1 explicitly penalizes visual hallucinations and shortcut guessing. By fundamentally shifting multimodal inference from black-box pattern matching to verifiable logical derivation, V-tableR1 4B establishes state-of-the-art accuracy among open-source models on complex tabular benchmarks, outperforming models up to 18x its size and improving over its SFT baseline