MM-Doc-R1: Training Agents for Long Document Visual Question Answering through Multi-turn Reinforcement Learning

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Lin, Jiahang, Hu, Kai, Wang, Binghai, Zhou, Yuhao, Xi, Zhiheng, Guo, Honglin, Liu, Shichun, Wang, Junzhe, Dou, Shihan, Zhou, Enyu, Yan, Hang, Han, Zhenhua, Gui, Tao, Zhang, Qi, Huang, Xuanjing
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908965226414080
author Lin, Jiahang
Hu, Kai
Wang, Binghai
Zhou, Yuhao
Xi, Zhiheng
Guo, Honglin
Liu, Shichun
Wang, Junzhe
Dou, Shihan
Zhou, Enyu
Yan, Hang
Han, Zhenhua
Gui, Tao
Zhang, Qi
Huang, Xuanjing
author_facet Lin, Jiahang
Hu, Kai
Wang, Binghai
Zhou, Yuhao
Xi, Zhiheng
Guo, Honglin
Liu, Shichun
Wang, Junzhe
Dou, Shihan
Zhou, Enyu
Yan, Hang
Han, Zhenhua
Gui, Tao
Zhang, Qi
Huang, Xuanjing
contents Conventional Retrieval-Augmented Generation (RAG) systems often struggle with complex multi-hop queries over long documents due to their single-pass retrieval. We introduce MM-Doc-R1, a novel framework that employs an agentic, vision-aware workflow to address long document visual question answering through iterative information discovery and synthesis. To incentivize the information seeking capabilities of our agents, we propose Similarity-based Policy Optimization (SPO), addressing baseline estimation bias in existing multi-turn reinforcement learning (RL) algorithms like GRPO. Our core insight is that in multi-turn RL, the more semantically similar two trajectories are, the more accurate their shared baseline estimation becomes. Leveraging this, SPO calculates a more precise baseline by similarity-weighted averaging of rewards across multiple trajectories, unlike GRPO which inappropriately applies the initial state's baseline to all intermediate states. This provides a more stable and accurate learning signal for our agents, leading to superior training performance that surpasses GRPO. Our experiments on the MMLongbench-Doc benchmark show that MM-Doc-R1 outperforms previous baselines by 10.4%. Furthermore, SPO demonstrates superior performance over GRPO, boosting results by 5.0% with Qwen3-8B and 6.1% with Qwen3-4B. These results highlight the effectiveness of our integrated framework and novel training algorithm in advancing the state-of-the-art for complex, long-document visual question answering.
format Preprint
id arxiv_https___arxiv_org_abs_2604_13579
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MM-Doc-R1: Training Agents for Long Document Visual Question Answering through Multi-turn Reinforcement Learning
Lin, Jiahang
Hu, Kai
Wang, Binghai
Zhou, Yuhao
Xi, Zhiheng
Guo, Honglin
Liu, Shichun
Wang, Junzhe
Dou, Shihan
Zhou, Enyu
Yan, Hang
Han, Zhenhua
Gui, Tao
Zhang, Qi
Huang, Xuanjing
Computation and Language
Conventional Retrieval-Augmented Generation (RAG) systems often struggle with complex multi-hop queries over long documents due to their single-pass retrieval. We introduce MM-Doc-R1, a novel framework that employs an agentic, vision-aware workflow to address long document visual question answering through iterative information discovery and synthesis. To incentivize the information seeking capabilities of our agents, we propose Similarity-based Policy Optimization (SPO), addressing baseline estimation bias in existing multi-turn reinforcement learning (RL) algorithms like GRPO. Our core insight is that in multi-turn RL, the more semantically similar two trajectories are, the more accurate their shared baseline estimation becomes. Leveraging this, SPO calculates a more precise baseline by similarity-weighted averaging of rewards across multiple trajectories, unlike GRPO which inappropriately applies the initial state's baseline to all intermediate states. This provides a more stable and accurate learning signal for our agents, leading to superior training performance that surpasses GRPO. Our experiments on the MMLongbench-Doc benchmark show that MM-Doc-R1 outperforms previous baselines by 10.4%. Furthermore, SPO demonstrates superior performance over GRPO, boosting results by 5.0% with Qwen3-8B and 6.1% with Qwen3-4B. These results highlight the effectiveness of our integrated framework and novel training algorithm in advancing the state-of-the-art for complex, long-document visual question answering.
title MM-Doc-R1: Training Agents for Long Document Visual Question Answering through Multi-turn Reinforcement Learning
topic Computation and Language
url https://arxiv.org/abs/2604.13579