arXiv cs.AI· Hangeol Chang, Changsun Lee, Seungjoon Rho, Junho Yeo, Jong Chul Ye·· 10 小时前AI 评分29
HYPER:面向证据检索的假设引导规划框架
HYPER: Hypothesis-Guided Planning for Evidence-Seeking Retrieval
AI 导读
HYPER 是一个无需训练的前置检索规划框架,将 RAG 从主题导向检索转向证据导向检索,以临时答案假设为规划锚点,把检索拆分为支持、区分、特征三种证据角色。在 MedQA 和 MMLU-Med 上,其平均准确率在十个基线中最佳,分别比 Simple RAG 提升 5.9 和 3.6 个百分点,增益也迁移到 MMLU-Pro 的 14 个学科,代价是每题多两次 LLM 调用。
正文 · 原文
Abstract:Retrieval-Augmented Generation (RAG) improves Large Language Models (LLMs) by grounding generation in external, non-parametric knowledge. However, when a task requires choosing among competing possibilities, such as a diagnosis among plausible alternatives, topically relevant context is often insufficient to drive the final decision. To address this issue, we propose HYPER: Hypothesis-Guided Planning for Evidence-Seeking Retrieval, a training-free pre-retrieval planning framework that reorients RAG from topic-oriented retrieval to evidence-oriented retrieval. HYPER uses a provisional answer hypothesis as a planning anchor. It then decomposes retrieval into three evidence roles. Support seeks evidence for the provisional answer. Distinction retrieves criteria that separate competing alternatives. Feature grounds salient question features needed for the decision. This planning step steers retrieval toward evidence near the decision boundary, enabling the generator to answer from context that is more directly aligned with the required choice. On MedQA and MMLU-Med, where single-query retrieval falls below no-retrieval chain-of-thought prompting, HYPER achieves the best average accuracy among ten baselines, improving over Simple RAG by 5.9 and 3.6 points, respectively. Its gains transfer to the 14 subjects of MMLU-Pro, at the cost of two additional LLM calls per question. Code is available at this https URL.
| Comments: | EACL 2027(Findings) |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2603.19008 [cs.CL] |
| (or arXiv:2603.19008v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2603.19008 arXiv-issued DOI via DataCite |
Submission history
From: Jong Chul Ye [view email]
[v1]
Thu, 19 Mar 2026 15:15:58 UTC (281 KB)
[v2]
Wed, 18 Nov 2026 12:02:07 UTC (132 KB)
来源:arXiv cs.AI · arxiv.org