Chart-RL: Generalized Chart Comprehension via Reinforcement Learning with Verifiable Rewards

TL;DR

Chart-RL introduces a reinforcement learning approach with verifiable rewards to improve chart question answering, demonstrating superior performance and robustness over existing models, and highlighting the importance of task complexity over data quantity.

Contribution

The paper presents a novel RL-based method with verifiable rewards for chart comprehension, enhancing generalization and reasoning in vision-language models.

Findings

Outperforms supervised fine-tuning on multiple benchmarks.

Achieves robustness across various perturbed chart categories.

Shows task complexity impacts learning more than data quantity.

Abstract

Accurate chart comprehension represents a critical challenge in advancing multimodal learning systems, as extensive information is compressed into structured visual representations. However, existing vision-language models (VLMs) frequently struggle to generalize on unseen charts because it requires abstract, symbolic, and quantitative reasoning over structured visual representations. In this work, we introduce Chart-RL, an effective reinforcement learning (RL) method that employs mathematically verifiable rewards to enhance chart question answering in VLMs. Our experiments demonstrate that Chart-RL consistently outperforms supervised fine-tuning (SFT) across different chart understanding benchmarks, achieving relative improvements of 16.7% on MutlChartQA, and 11.5% on ChartInsights. We conduct robustness analysis, where Chart-RL achieves enhanced performance in 18 of 25 perturbed chart categories, demonstrating strong consistency and reasoning capability across visual variations. Furthermore, we demonstrate that task difficulty and inherent complexity are more critical than data quantity in RL training. For instance, Chart-RL trained on merely 10 complex chart-query examples significantly outperforms models trained on over 6,000 simple examples. Additionally, training on challenging reasoning tasks not only improves in-domain generalization relative to simpler tasks, but also facilitate strong transfer to out-of-domain visual mathematical problems.