arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2601.04651cs.AIcs.IRcs.MA
收藏

对抗 yet 合作:检索增强语言模型中的多视角推理

Adversarial Yet Cooperative: Multi-Perspective Reasoning in Retrieved-Augmented Language Models

East China Normal University(华东师范大学);Baidu Inc.(百度公司);Southeast University(东南大学)

Can Xu, Lingyong Yan, Jiayi Wu, Haosen Wang, Shuaiqiang Wang, Yuchen Li, Jizhou Huang, Dawei Yin, Xiang Li

AI总结:

本文提出对抗推理RAG框架,通过推理与验证的对抗机制提升检索增强语言模型的多视角推理能力与验证严谨性。

摘要:

近年来,将大型推理模型(LRMs)与检索增强生成(RAG)相结合的进展显示出了有希望的结果,但仍然存在两个关键挑战:(1)推理模型通常从单一、无争议的视角进行操作,限制了它们在外部文档上进行深入、自我纠正的推理能力;(2)现有训练范式过度依赖以结果为导向的奖励,这为塑造复杂的多步骤推理过程提供了不足的信号。为了解决这些问题,我们提出了一个名为对抗推理RAG(ARR)的推理-验证框架。推理器和验证器在检索到的证据上进行推理,并相互批判逻辑,同时受到过程感知优势的指导,该奖励不需要外部评分模型。这种奖励结合了显式的观察信号和内部模型不确定性,以共同优化推理的准确性和验证的严谨性。在多个基准测试上的实验展示了我们方法的有效性。

英文摘要:

Recent advances in synergizing large reasoning models (LRMs) with retrieval-augmented generation (RAG) have shown promising results, yet two critical challenges remain: (1) reasoning models typically operate from a single, unchallenged perspective, limiting their ability to conduct deep, self-correcting reasoning over external documents, and (2) existing training paradigms rely excessively on outcome-oriented rewards, which provide insufficient signal for shaping the complex, multi-step reasoning process. To address these issues, we propose an Reasoner-Verifier framework named Adversarial Reasoning RAG (ARR). The Reasoner and Verifier engage in reasoning on retrieved evidence and critiquing each other's logic while being guided by process-aware advantage that requires no external scoring model. This reward combines explicit observational signals with internal model uncertainty to jointly optimize reasoning fidelity and verification rigor. Experiments on multiple benchmarks demonstrate the effectiveness of our method.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑