Chart-RVR: Reinforcement Learning with Verifiable Rewards for Explainable Chart Reasoning
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914089694920704 |
|---|---|
| author | Sinha, Sanchit Frunza, Oana Rasul, Kashif Nevmyvaka, Yuriy Zhang, Aidong |
| author_facet | Sinha, Sanchit Frunza, Oana Rasul, Kashif Nevmyvaka, Yuriy Zhang, Aidong |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_10973 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Chart-RVR: Reinforcement Learning with Verifiable Rewards for Explainable Chart Reasoning Sinha, Sanchit Frunza, Oana Rasul, Kashif Nevmyvaka, Yuriy Zhang, Aidong Computer Vision and Pattern Recognition Machine Learning 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. |
| title | Chart-RVR: Reinforcement Learning with Verifiable Rewards for Explainable Chart Reasoning |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2510.10973 |