arXiv cs.AI· Lihong He, Lucian Popa, Marina Danilevsky·· 13 小时前AI 评分31
ClariEdge：轻量微调 LoRA 让小型模型更好识别 RAG 模糊查询并生成澄清请求
A Small Model Asks Better: Fine-Tuning Query Clarification for Retrieval-Augmented Generation
AI 导读
研究提出 ClariEdge 数据集，约 3900 个样本，覆盖多轮对话中需澄清的模糊查询及无需澄清的负例。轻量微调 LoRA 在该任务上达到 96.6% 分类准确率，在分类与澄清生成质量上均优于多个更大模型的提示词基线。硬负例是区分真正模糊与看似模糊查询的关键，该 LoRA 在检索前和检索后场景均表现良好，可在 RAG 流程中两处用作答案生成前的检查。
正文 · 原文
Abstract:Query clarification is essential for retrieval-augmented generation (RAG) systems that face ambiguous user queries. We consider multi-turn grounded conversational scenarios where a user query is either ill-formed or ambiguous in the context of the prior dialog or, more interestingly, it may be well-formed but when posed against retrieved documents there are multiple relevant interpretations. The task is to recognize both such cases and then generate an appropriate clarification request to the user. We present ClariEdge, a new dataset of approximately 3,900 examples that include such ambiguous queries along with negative examples that do not require clarification. We also show that a lightweight fine-tuned LoRA achieves 96.6% classification accuracy and outperforms prompted baselines of many larger models on both classification and quality of clarification generation. Our study shows that hard negatives are an essential ingredient to separate out queries that are truly ambiguous from queries that may look ambiguous but are not (i.e., can be resolved from the context). The same LoRA achieves high performance in both pre-retrieval and post-retrieval scenarios; thus, it can be used in two places in a RAG pipeline as a check before answer generation. The ClariEdge dataset is available at this https URL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.28686 [cs.AI] |
| (or arXiv:2610.28686v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.28686 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Lihong He [view email]
[v1]
Mon, 19 Oct 2026 22:33:41 UTC (12,180 KB)
来源：arXiv cs.AI · arxiv.org