From Sight to Insight: Improving Visual Reasoning Capabilities of Multimodal Models via Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909979778220032 |
|---|---|
| author | Sharif, Omar Hossain, Eftekhar Ng, Patrick |
| author_facet | Sharif, Omar Hossain, Eftekhar Ng, Patrick |
| contents | Reinforcement learning (RL) has emerged as a promising approach for eliciting reasoning chains before generating final answers. However, multimodal large language models (MLLMs) generate reasoning that lacks integration of visual information. This limits their ability to solve problems that demand accurate visual perception, such as visual puzzles. We show that visual perception is the key bottleneck in such tasks: converting images into textual descriptions significantly improves performance, yielding gains of 26.7% for Claude 3.5 and 23.6% for Claude 3.7.
To address this, we investigate reward-driven RL as a mechanism to unlock long visual reasoning in open-source MLLMs without requiring costly supervision. We design and evaluate six reward functions targeting different reasoning aspects, including image understanding, thinking steps, and answer accuracy. Using group relative policy optimization (GRPO), our approach explicitly incentivizes longer, structured reasoning and mitigates bypassing of visual information. Experiments on Qwen-2.5-VL-7B achieve 5.56% improvements over the base model, with consistent gains across both in-domain and out-of-domain settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_00215 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | From Sight to Insight: Improving Visual Reasoning Capabilities of Multimodal Models via Reinforcement Learning Sharif, Omar Hossain, Eftekhar Ng, Patrick Computer Vision and Pattern Recognition Computation and Language Reinforcement learning (RL) has emerged as a promising approach for eliciting reasoning chains before generating final answers. However, multimodal large language models (MLLMs) generate reasoning that lacks integration of visual information. This limits their ability to solve problems that demand accurate visual perception, such as visual puzzles. We show that visual perception is the key bottleneck in such tasks: converting images into textual descriptions significantly improves performance, yielding gains of 26.7% for Claude 3.5 and 23.6% for Claude 3.7. To address this, we investigate reward-driven RL as a mechanism to unlock long visual reasoning in open-source MLLMs without requiring costly supervision. We design and evaluate six reward functions targeting different reasoning aspects, including image understanding, thinking steps, and answer accuracy. Using group relative policy optimization (GRPO), our approach explicitly incentivizes longer, structured reasoning and mitigates bypassing of visual information. Experiments on Qwen-2.5-VL-7B achieve 5.56% improvements over the base model, with consistent gains across both in-domain and out-of-domain settings. |
| title | From Sight to Insight: Improving Visual Reasoning Capabilities of Multimodal Models via Reinforcement Learning |
| topic | Computer Vision and Pattern Recognition Computation and Language |
| url | https://arxiv.org/abs/2601.00215 |