KunLunBaizeRAG: Reinforcement Learning Driven Inference Performance Leap for Large Language Models

Cheng Li; Jiexiong Liu; Yixuan Chen; Qihang Zhou; KunLun Meta

KunLunBaizeRAG: Reinforcement Learning Driven Inference Performance Leap for Large Language Models

Artificial Intelligence 2025-06-30 v2

Authors: Cheng Li , Jiexiong Liu , Yixuan Chen , Qihang Zhou , KunLun Meta

Abstract

This paper introduces KunLunBaizeRAG, a reinforcement learning-driven reasoning framework designed to enhance the reasoning capabilities of large language models (LLMs) in complex multi-hop question-answering tasks. The framework addresses key limitations of traditional RAG, such as retrieval drift, information redundancy, and strategy rigidity. Key innovations include the RAG-driven Reasoning Alignment (RDRA) mechanism, the Search-Think Iterative Enhancement (STIE) mechanism, the Network-Local Intelligent Routing (NLR) mechanism, and a progressive hybrid training strategy. Experimental results demonstrate significant improvements in exact match (EM) and LLM-judged score (LJ) across four benchmarks, highlighting the framework's robustness and effectiveness in complex reasoning scenarios.

Keywords

retrieval-augmented generation automated reasoning reinforcement learning

Cite

@article{arxiv.2506.19466,
  title  = {KunLunBaizeRAG: Reinforcement Learning Driven Inference Performance Leap for Large Language Models},
  author = {Cheng Li and Jiexiong Liu and Yixuan Chen and Qihang Zhou and KunLun Meta},
  journal= {arXiv preprint arXiv:2506.19466},
  year   = {2025}
}

KunLunBaizeRAG: Reinforcement Learning Driven Inference Performance Leap for Large Language Models

Abstract

Keywords

Cite

Related papers