arXiv cs.AI· Chaolang Li, Jingyu Li, Zhaocheng Du, Xuyang Xie, Gang Wang, Pengwen Dai·· 13 小时前AI 评分31
NHNM：从正例中挖掘自然硬负例，提升指令跟随检索
Natural Hard-Negative Mining from Positives for Instruction-Following Retrieval
AI 导读
研究者提出自然硬负例挖掘方法 NHNM，将原始正例段落改造成符合查询指令的自然硬负例，通过候选清洗、反转构造与几何过滤自动生成高对比监督信号。在 MSMARCO 上构建出约 7K 条实例的 IFIR 训练集 NHNM-MARCO，568M 至 4B 参数模型微调后在 InstructIR 和 FollowIR 上表现强劲，部分设置下可匹敌或超越更大规模检索模型。
正文 · 原文
Abstract:Instruction-following information retrieval (IFIR) requires retrievers to rank documents by both a query and its instance-specific instructions, yet instruction-conditioned contrastive supervision remains scarce. Existing synthetic methods typically fix a positive passage and generate additional negatives, overlooking the hard-negative potential of the original positive. We propose Natural Hard-Negative Mining (NHNM), which promotes convertible negatives under query-compatible instructions and repurposes original positives as natural instruction hard negatives. Its automated Candidate Cleaning, Reversal Construction, and Geometry Filtering pipeline produces compact, high-contrast supervision aligned with the pretrained retrieval space. Applied to MSMARCO, NHNM yields NHNM-MARCO, an IFIR training set of nearly 7K instances. On InstructIR and FollowIR, models with 568M to 4B parameters fine-tuned on NHNM-MARCO achieve strong instruction-following performance and match or surpass substantially larger retrievers in several settings. Controlled analyses and ablations attribute these gains to natural hard-negative supervision rather than model scale or training-data volume.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.36900 [cs.AI] |
| (or arXiv:2610.36900v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36900 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Chaolang Li [view email]
[v1]
Tue, 27 Oct 2026 09:31:15 UTC (1,656 KB)
来源：arXiv cs.AI · arxiv.org