Guardado en:
Detalles Bibliográficos
Autores principales: Liu, Yu, Zhang, Wenxiao, Cao, Cong, Lu, Wenxuan, Yuan, Fangfang, Guo, Diandian, Peng, Kun, Sun, Qiang, Zhang, Kaiyan, Liu, Yanbing, Hong, Jin B., Zhou, Bowen, Ma, Zhiyuan
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2601.05465
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911362803826688
author Liu, Yu
Zhang, Wenxiao
Cao, Cong
Lu, Wenxuan
Yuan, Fangfang
Guo, Diandian
Peng, Kun
Sun, Qiang
Zhang, Kaiyan
Liu, Yanbing
Hong, Jin B.
Zhou, Bowen
Ma, Zhiyuan
author_facet Liu, Yu
Zhang, Wenxiao
Cao, Cong
Lu, Wenxuan
Yuan, Fangfang
Guo, Diandian
Peng, Kun
Sun, Qiang
Zhang, Kaiyan
Liu, Yanbing
Hong, Jin B.
Zhou, Bowen
Ma, Zhiyuan
contents Answering real-world open-domain multi-hop questions over massive corpora is a critical challenge in Retrieval-Augmented Generation (RAG) systems. Recent research employs reinforcement learning (RL) to end-to-end optimize the retrieval-augmented reasoning process, directly enhancing its capacity to resolve complex queries. However, reliable deployment is hindered by two obstacles. 1) Retrieval Collapse: iterative retrieval over large corpora fails to locate intermediate evidence containing bridge answers without reasoning-guided planning, causing downstream reasoning to collapse. 2) Learning Instability: end-to-end trajectory training suffers from weak credit assignment across reasoning chains and poor error localization across modules, causing overfitting to benchmark-specific heuristics that limit transferability and stability. To address these problems, we propose PRISMA, a decoupled RL-guided framework featuring a Plan-Retrieve-Inspect-Solve-Memoize architecture. PRISMA's strength lies in reasoning-guided collaboration: the Inspector provides reasoning-based feedback to refine the Planner's decomposition and fine-grained retrieval, while enforcing evidence-grounded reasoning in the Solver. We optimize individual agent capabilities via Two-Stage Group Relative Policy Optimization (GRPO). Stage I calibrates the Planner and Solver as specialized experts in planning and reasoning, while Stage II utilizes Observation-Aware Residual Policy Optimization (OARPO) to enhance the Inspector's ability to verify context and trigger targeted recovery. Experiments show that PRISMA achieves state-of-the-art performance on ten benchmarks and can be deployed efficiently in real-world scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2601_05465
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PRISMA: Reinforcement Learning Guided Two-Stage Policy Optimization in Multi-Agent Architecture for Open-Domain Multi-Hop Question Answering
Liu, Yu
Zhang, Wenxiao
Cao, Cong
Lu, Wenxuan
Yuan, Fangfang
Guo, Diandian
Peng, Kun
Sun, Qiang
Zhang, Kaiyan
Liu, Yanbing
Hong, Jin B.
Zhou, Bowen
Ma, Zhiyuan
Artificial Intelligence
Answering real-world open-domain multi-hop questions over massive corpora is a critical challenge in Retrieval-Augmented Generation (RAG) systems. Recent research employs reinforcement learning (RL) to end-to-end optimize the retrieval-augmented reasoning process, directly enhancing its capacity to resolve complex queries. However, reliable deployment is hindered by two obstacles. 1) Retrieval Collapse: iterative retrieval over large corpora fails to locate intermediate evidence containing bridge answers without reasoning-guided planning, causing downstream reasoning to collapse. 2) Learning Instability: end-to-end trajectory training suffers from weak credit assignment across reasoning chains and poor error localization across modules, causing overfitting to benchmark-specific heuristics that limit transferability and stability. To address these problems, we propose PRISMA, a decoupled RL-guided framework featuring a Plan-Retrieve-Inspect-Solve-Memoize architecture. PRISMA's strength lies in reasoning-guided collaboration: the Inspector provides reasoning-based feedback to refine the Planner's decomposition and fine-grained retrieval, while enforcing evidence-grounded reasoning in the Solver. We optimize individual agent capabilities via Two-Stage Group Relative Policy Optimization (GRPO). Stage I calibrates the Planner and Solver as specialized experts in planning and reasoning, while Stage II utilizes Observation-Aware Residual Policy Optimization (OARPO) to enhance the Inspector's ability to verify context and trigger targeted recovery. Experiments show that PRISMA achieves state-of-the-art performance on ten benchmarks and can be deployed efficiently in real-world scenarios.
title PRISMA: Reinforcement Learning Guided Two-Stage Policy Optimization in Multi-Agent Architecture for Open-Domain Multi-Hop Question Answering
topic Artificial Intelligence
url https://arxiv.org/abs/2601.05465