The paper introduces Natural Hard-Negative Mining (NHNM) for instruction‑following information retrieval, a method that turns original positive passages into hard negatives under query‑compatible instructions. NHNM uses a pipeline of Candidate Cleaning, Reversal Construction, and Geometry Filtering to generate a compact, high‑contrast training set called NHNM‑MARCO with nearly 7,000 instances. Models fine‑tuned on NHNM‑MARCO, ranging from 568 M to 4 B parameters, achieve strong instruction‑following performance on InstructIR and FollowIR, matching or exceeding larger retrievers and demonstrating that the gains stem from natural hard‑negative supervision rather than model size or data volume.
By Chaolang Li, Jingyu Li, Zhaocheng Du, Xuyang Xie, Gang Wang, Pengwen Dai
arXiv:2606. 01304v1 Announce Type: new Abstract: Hard negative mining has become the dominant strategy for training retrievers, yet it faces intrinsic limitations: negatives are bounded by corpus availability, selected by retriever score rather than diagnostic value, and increasingly contaminated by false positives as the retriever improves.
By Zhicheng Zhang, Jiwei Tang, Kuicai Dong, Xiaopeng Li, Jieming Zhu, Jingyu Li, Qianhui Zhu, Fengyuan Lu, Wang Jiaheng, Gang Wang, Hai-Tao Zheng, Zhaocheng Du
The paper investigates modular entity disambiguation by separating candidate retrieval from entity selection. It compares sparse retrieval (BM25), Web KB search, and a dense retriever, all paired with large language model selectors. Results show that a training‑free BM25 retriever combined with an LLM selector achieves state‑of‑the‑art performance on the ZELDA benchmark, and the modular approach enables abstention when retrieval fails.
By Fina Polat, Daniel Daza, Pengyu Zhang, Klim Zaporojets, Paul Groth
arXiv:2609.22227v1 Announce Type: cross
Abstract: Generative retrieval represents each item by a short Semantic ID and casts recommendation as autoregressive generation of that sequence. Because the...
By Bin Wang, Zhengyu Zhang
E‑SENS is a training‑free reranking technique designed to improve negative‑constraint retrieval in retrieval‑augmented language models. It works by extracting a compact ‘trap query’ that represents the excluded concept and subtracting the similarity of this trap query from the original retrieval score, thereby reducing the influence of documents that mention the excluded concept. Experiments on the ExcluIR benchmark demonstrate that E‑SENS achieves a clear recall‑violation trade‑off across four embedding models and effectively reduces trap retrieval while preserving recall.
By Yerang Kim, Jiyoon Myung, Joohyung Han
arXiv:2603. 20990v3 Announce Type: replace-cross Abstract: Hard-negative source selection for dense retrieval is usually decided only after fine-tuning and downstream evaluation.
By Aarush Sinha, Rahul Seetharaman, Aman Bansal
The paper introduces Compositional Concept Unlearning (C$^2$U), a method for suppressing unwanted concepts in retrieval‑augmented generation by adding a bounded edited‑minus‑original delta to the raw query. Using a fixed encoder and document index, C$^2$U routes a profile‑derived direction through an augmented atom bank, achieving a 4.25‑percentage‑point reduction in pooled binary target Hit@5 across five datasets, four encoders, and three seeds without increasing control‑query exposure. The approach also outperforms an averaged‑direction comparator in a four‑concept composition setting after independent validation gates and first‑use test evaluation.
By Yiyang Duan, Fan Wu, Tiantong Wu, Che Wang, Zhiren Gong, Wei Guo, Yang Cao, Wei Yang Bryan Lim
arXiv:2607. 18785v2 Announce Type: replace Abstract: As large language model agents gain access to increasingly large skill libraries, retrieving the right skill becomes critical to reliable capability selection and execution.
By Jinying Xiao, Bin Li, Xiaopeng Li, Jianling Li, Jiacheng Jie, Xiaodong Liu, Ma Jun, Chao Wang, Nyima Tashi, Jie Yu
arXiv:2606. 03056v1 Announce Type: new Abstract: As LLM agents adopt large skill libraries, selecting the right subset becomes a structural problem rather than a similarity-matching one: skills depend on, conflict with, specialize, or duplicate one another, a structure invisible to both full enumeration and embedding similarity.
By Tong Bai, Zhenglin Wan, Pengfei Zhou, Xingrui Yu, Wangbo Zhao, Yang You, Ivor W. Tsang
arXiv:2610.02516v1 Announce Type: cross
Abstract: Zero-shot classifiers are useful for routing user requests to specialized LLM tasks, but scoring every request against a large candidate set is expen...
By Haifeng Wu, Srinivasan Manoharan, Jian Wan, Fangbo Tu, Junhua Zhao, Xin Chen
arXiv:2608. 15445v1 Announce Type: new Abstract: When a reward is correct on every training example yet consistent with more than one goal, a model can acquire an unintended one, a failure known as goal misgeneralization.
By Suyash Maniyar, Armaan Sandhu, Abhishek Mishra
The paper introduces DEPT, a method that trains a single decoder-only large language model to both expand queries and encode documents for retrieval. By preserving document embeddings close to their initial cached values while allowing gradients to flow through the generator, DEPT stabilizes retrieval targets and enables efficient index reuse and online hard‑negative mining. Experiments on the BEIR benchmark with Qwen3‑4B‑Instruct‑2507 and LLaMA‑3.2‑3B‑Instruct show that DEPT outperforms training‑free, independently trained, and staged unified baselines, with ablations confirming the benefits of preservation, whitening, end‑to‑end expansion training, and online negatives.
By Jingyuan Wang, Richong Zhang, Zhijie Nie, Mingxin Li, Yanzhao Zhang