![[2602.12395] What does RL improve for Visual Reasoning? A Frankenstein-Style Analysis](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.12395] What does RL improve for Visual Reasoning? A Frankenstein-Style Analysis
Summary
This paper explores the impact of reinforcement learning (RL) on visual reasoning capabilities in vision-language models, proposing a novel analysis framework to assess improvements over traditional supervised fine-tuning.
Why It Matters
Understanding how RL enhances visual reasoning is crucial for developing more effective AI models. This research highlights the specific areas where RL contributes to model performance, which can guide future advancements in multimodal AI systems.
Key Takeaways
- RL improves visual reasoning by refining mid-to-late transformer computations.
- The proposed framework allows for a clearer attribution of improvements to specific skills.
- Benchmark evaluations may not fully capture the nuances of RL's contributions.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.12395 (cs) [Submitted on 12 Feb 2026] Title:What does RL improve for Visual Reasoning? A Frankenstein-Style Analysis Authors:Xirui Li, Ming Li, Tianyi Zhou View a PDF of the paper titled What does RL improve for Visual Reasoning? A Frankenstein-Style Analysis, by Xirui Li and 1 other authors View PDF HTML (experimental) Abstract:Reinforcement learning (RL) with verifiable rewards has become a standard post-training stage for boosting visual reasoning in vision-language models, yet it remains unclear what capabilities RL actually improves compared with supervised fine-tuning as cold-start initialization (IN). End-to-end benchmark gains conflate multiple factors, making it difficult to attribute improvements to specific skills. To bridge the gap, we propose a Frankenstein-style analysis framework including: (i) functional localization via causal probing; (ii) update characterization via parameter comparison; and (iii) transferability test via model merging. Instead, RL induces a consistent inference-time shift primarily in mid-to-late layers, and these mid-to-late refinements are both transferable (via merging) and necessary (via freezing) for RL gains. Overall, our results suggest that RL's reliable contribution in visual reasoning is not a uniform enhancement of visual perception, but a systematic refinement of mid-to-late transformer computation that improves vision-to-reasoning alignment and reasoning ...
