UniCAR-RL: Seeing Better before Thinking Deeper in Visual Mathematics

Researchers introduced UniCAR-RL, an annotation-free reinforcement learning framework that decouples perception and reasoning optimization in Multimodal Large Language Models (MLLMs) to improve complex mathematical visual reasoning, according to arXiv paper 2609.13849v1. UniCAR-RL uses three branches — Caption-RL for verifier-guided perception, Reasoning-RL for logic based on a gold image description, and QA-RL for end-to-end alignment — and substantially improves MLLM mathematical and visual reasoning using only raw short-answer data, with strong generalization across architectures and scales.

arXiv:2609.13849v1 Announce Type: new Abstract: Multimodal Large Language Models MLLMs often struggle with complex mathematical visual reasoning primarily due to a lack of fine-grained perception, causing initial visual hallucinations to directly trigger cascading reasoning failures. In traditional end-to-end reinforcement learning RL , sparse rewards fail to decouple perceptual hallucinations from logical missteps, hindering targeted perception optimization. Alternatively, fine-tuning with perception-enhanced CoT data incurs high costs and hallucinations. In this paper, we address these challenges by proposing UniCAR-RL, an annotation-free RL framework. By explicitly decoupling the optimization of perception and reasoning during the training process, it achieves isolation and optimization of both capabilities. Specifically, UniCAR-RL consists of three synergistic branches: 1 a Caption-RL branch that optimizes perception capabilities through verifier-guided reasoning validation; 2 a Reasoning-RL branch that performs logical reasoning based on a gold image description to halt cascading errors; 3 a QA-RL branch that retains native end-to-end alignment to ensure robust question-answering performance. Experiments show that UniCAR-RL substantially improves MLLMs' mathematical and visual reasoning using only raw short-answer data. Furthermore, it demonstrates strong generalization across diverse architectures and scales.