UniDoc-RL: Coarse-to-Fine Visual RAG with Hierarchical Actions and Dense Rewards

TL;DR

UniDoc-RL is a reinforcement learning framework that enhances visual reasoning in Large Vision-Language Models by hierarchically refining visual evidence and using dense rewards for improved performance.

Contribution

It introduces a hierarchical action space and dense multi-reward scheme for end-to-end training of visual retrieval and reasoning models.

Findings

Achieves up to 17.7% improvements over previous RL-based methods.

Effectively refines visual evidence from coarse retrieval to fine-grained perception.

Surpasses state-of-the-art baselines on three benchmark datasets.

Abstract

Retrieval-Augmented Generation (RAG) extends Large Vision-Language Models (LVLMs) with external visual knowledge. However, existing visual RAG systems typically rely on generic retrieval signals that overlook the fine-grained visual semantics essential for complex reasoning. To address this limitation, we propose UniDoc-RL, a unified reinforcement learning framework in which an LVLM agent jointly performs retrieval, reranking, active visual perception, and reasoning. UniDoc-RL formulates visual information acquisition as a sequential decision-making problem with a hierarchical action space. Specifically, it progressively refines visual evidence from coarse-grained document retrieval to fine-grained image selection and active region cropping, allowing the model to suppress irrelevant content and attend to information-dense regions. For effective end-to-end training, we introduce a dense multi-reward scheme that provides task-aware supervision for each action. Based on Group Relative Policy Optimization (GRPO), UniDoc-RL aligns agent behavior with multiple objectives without relying on a separate value network. To support this training paradigm, we curate a comprehensive dataset of high-quality reasoning trajectories with fine-grained action annotations. Experiments on three benchmarks demonstrate that UniDoc-RL consistently surpasses state-of-the-art baselines, yielding up to 17.7% gains over prior RL-based methods.