Chart-RVR: Reinforcement Learning with Verifiable Rewards for Explainable Chart Reasoning

2025-10-12

AI Summary

Large Vision-Language Models (LVLMs) suffer from weak out-of-distribution (OOD) generalization and unreliable chain-of-thought (CoT) reasoning in chart understanding, undermining interpretability and robustness. To address these challenges, we propose Chart-RVR—a novel framework that jointly optimizes chart type recognition, table reconstruction, and reasoning generation. Chart-RVR introduces Group-wise Relative Policy Optimization (GRPO) coupled with a triple automatic verifiable reward mechanism, enforcing concurrent constraints on CoT faithfulness, structural consistency, and logical verifiability. Evaluated on a 3B-parameter LVLM across six chart reasoning benchmarks, Chart-RVR achieves state-of-the-art performance, substantially narrows the in-distribution vs. OOD accuracy gap, and generates more accurate, interpretable, and verifiable reasoning paths. To our knowledge, this is the first work to systematically enhance both generalization capability and interpretability of LVLMs for chart understanding tasks.

Technology Category

• Computer Vision: Visual Reasoning & Symbolic Representations
• Knowledge Representation and Reasoning: Common-Sense Reasoning
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

The capabilities of Large Vision-Language Models (LVLMs) have reached state-of-the-art on many visual reasoning tasks, including chart reasoning, yet they still falter on out-of-distribution (OOD) data, and degrade further when asked to produce their chain-of-thought (CoT) rationales, limiting explainability. We present Chart-RVR, a general framework that fine-tunes LVLMs to be more robust and explainable for chart reasoning by coupling Group Relative Policy Optimization (GRPO) with automatically verifiable rewards. Our framework comprises of three rewards that maximize: (i) correct chart-type classification, (ii) faithful chart table reconstruction, and (iii) process conformity. Applied to 3-billion-parameter LVLMs, Chart-RVR consistently outperforms standard supervised fine-tuning (SFT) on both in-distribution and out-of-distribution datasets, closing the OOD performance gap while improving rationale fidelity. The resulting models, the Chart-RVR-3B series, achieve state-of-the-art results on six chart-reasoning benchmarks spanning in-domain and OOD settings, surpassing all existing models of comparable size. Beyond accuracy, Chart-RVR yields more interpretable CoT rationales, strengthening trust and reliability - showcasing the power of verifiable rewards with GRPO for training reliable, interpretable chart-reasoning models.

Problem

• Improving chart reasoning robustness on out-of-distribution data
• Enhancing explainability through faithful chain-of-thought rationales
• Developing verifiable reward framework for reliable chart interpretation

Innovation

• Fine-tunes LVLMs with Group Relative Policy Optimization
• Uses verifiable rewards for chart-type classification
• Improves chart table reconstruction and process conformity