AgenticRag-R1 Framework for Multi-Step Reasoning
AgenticRag-R1: Agentic Reinforcement Learning with Stack Memory for Multi-Step Reasoning, Retrieval and Memorizing
AgenticRag-R1 improves long-horizon reasoning tasks relevant to developers working with complex RAG systems in ASEAN's growing AI sector.
AgenticRag-R1 introduces a reinforcement learning framework that integrates reasoning, retrieval, and memory through a memory stack and fine-grained action space.
The model uses hierarchical action-aware rewards and information-aware trajectory rejection to enable long-horizon learning.
Experiments across multi-hop, open-domain, and agentic reasoning benchmarks show consistent outperformance of strong baselines.
The framework learns more robust, interpretable, and memory-aware reasoning behaviors.
AgenticRag-R1: Agentic Reinforcement Learning with Stack Memory for Multi-Step Reasoning, Retrieval and Memorizing
Retrieval-Augmented Generation (RAG) improves the factuality of large language models (LLMs), yet existing RAG systems often struggle with complex, multi-step reasoning that requires adaptive retrieval and continuous revision of intermediate contexts. Recent reinforcement learning (RL)-based agentic RAG methods partially alleviate this issue, but typically rely on coarse-grained action spaces and trajectory-level rewards, resulting in weak reward assignment and a bias toward short-horizon, stereotyped reasoning template. To address, we propose AgenticRag-R1, a RL framework that deeply integrates reasoning, retrieval, and memory via a memory stack and fine-grained action space, supported by hierarchical action-aware rewards and an information-aware trajectory rejection strategy to enable effective long-horizon learning. Experiments across a diverse set of multi-hop, open-domain, and agentic reasoning benchmarks, spanning multiple backbone model sizes, demonstrate that AgenticRag-R1 consistently outperforms strong baselines. Moreover, AgenticRag-R1 learns more robust, interpretable, and memory-aware reasoning behaviors, highlighting the effect of fine-grained action modeling and information-aware optimization for long-horizon reasoning. Our code is anonymous available at https://github.com/jiangxinke/Harness-RL/tree/AgenticRAG-R1-Whitebox.
This report is categorized under AITOP's "Papers" desk. Key entities and technical topics under continuous tracking include AgenticRag-R1, RAG, reinforcement learning, multi-step reasoning.
Source attribution: curated from arXiv cs.AI (@Xinke Jiang, Yue Fang, Zhibang Yang, Jiaran Gao, Zhixin Zhang, Tao Feng, Rihong Qiu, Wentao Zhang, Hongxin Ding, Ruizhe Zhang, Yongxin Xu, Yuheng Huang, Xu Chu, Junfeng Zhao, Yasha Wang).