CoChat
Sign Up
Cited by 248Information Retrieval and Search BehaviorTopic ModelingAdvanced Image and Video Retrieval Techniques

Optimizing Dense Retrieval Model Training with Hard Negatives

Open access
Zhan, Jingtao, Mao, Jiaxin, Liu, Yiqun, Guo, Jiafeng, Zhang, Min, Ma, Shaoping·2021·Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval
DOI 10.1145/3404835.3462880
Open this paper in CoChat — free accountOpen PDF

Abstract

Ranking has always been one of the top concerns in information retrieval researches. For decades, the lexical matching signal has dominated the ad-hoc retrieval process, but solely using this signal in retrieval may cause the vocabulary mismatch problem. In recent years, with the development of representation learning techniques, many researchers turn to Dense Retrieval (DR) models for better ranking performance. Although several existing DR models have already obtained promising results, their performance improvement heavily relies on the sampling of training examples. Many effective sampling strategies are not efficient enough for practical usage, and for most of them, there still lacks theoretical analysis in how and why performance improvement happens. To shed light on these research questions, we theoretically investigate different training strategies for DR models and try to explain why hard negative sampling performs better than random sampling. Through the analysis, we also find that there are many potential risks in static hard negative sampling, which is employed by many existing training methods. Therefore, we propose two training strategies named a Stable Training Algorithm for dense Retrieval (STAR) and a query-side training Algorithm for Directly Optimizing Ranking pErformance (ADORE), respectively. STAR improves the stability of DR training process by introducing random negatives. ADORE replaces the widely-adopted static hard negative sampling method with a dynamic one to directly optimize the ranking performance. Experimental results on two publicly available retrieval benchmark datasets show that either strategy gains significant improvements over existing competitive baselines and a combination of them leads to the best performance.

Integrity signals

Registry

Crossref · 10.1145/3404835.3462880

Cites

21 of 21 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Recent studies on naive RAG often prefer the use of neural retrieval models over traditional term-matching models.

    “recent studies on naive RAG often prefer the use of neural retrieval models (e.g., dense retrieval models [96]) over traditional term-matching models (e.g., BM25 [97]).”

    From Foundations of GenIR (2025) · cites this paper as [96]

Cites

Flagged first, then by citations

Research works printed but not linked: 16

  1. Payal Bajaj, Daniel Campos, Nick Craswell, Li Deng, Jianfeng Gao, Xiaodong Liu, Rangan Majumder, Andrew McNamara, Bhaskar Mitra, Tri Nguyen, et almbox. 2016. Ms marco: A human generated machine reading comprehension dataset. arXiv preprint arXiv:1611.09268 (2016). · arXiv
  2. Yoshua Bengio, Aaron Courville, and Pascal Vincent. 2013. Representation learning: A review and new perspectives. IEEE transactions on pattern analysis and machine intelligence , Vol. 35, 8 (2013), 1798--1828.
  3. Zhuyun Dai and Jamie Callan. 2019. Context-aware sentence/passage term importance estimation for first stage retrieval. arXiv preprint arXiv:1910.10687 (2019). · arXiv
  4. Yingqi Qu Yuchen Ding, Jing Liu, Kai Liu, Ruiyang Ren, Xin Zhao, Daxiang Dong, Hua Wu, and Haifeng Wang. 2020. RocketQA: An Optimized Training Approach to Dense Passage Retrieval for Open-Domain Question Answering. arXiv preprint arXiv:2010.08191 (2020). · arXiv
  5. Luyu Gao, Zhuyun Dai, Zhen Fan, and Jamie Callan. 2020. Complementing lexical retrieval with semantic residual embedding. arXiv preprint arXiv:2004.13969 (2020). · arXiv
  6. Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Ming-Wei Chang. 2020. Realm: retrieval-augmented language model pre-training. arXiv preprint arXiv:2002.08909 (2020). · arXiv
  7. Shuguang Han, Xuanhui Wang, Mike Bendersky, and Marc Najork. 2020. Learning-to-Rank with BERT in TF-Ranking. arXiv preprint arXiv:2004.08476 (2020). · arXiv
  8. Vladimir Karpukhin, Barlas Oug uz, Sewon Min, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. arXiv preprint arXiv:2004.04906 (2020). · arXiv
  9. Sheng-Chieh Lin, Jheng-Hong Yang, and Jimmy Lin. 2020. Distilling Dense Representations for Ranking using Tightly-Coupled Teachers. arXiv preprint arXiv:2010.11386 (2020). · arXiv
  10. Tie-Yan Liu. 2009. Learning to rank for information retrieval. Foundations and trends in information retrieval , Vol. 3, 3 (2009), 225--331.
  11. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. RoBERTa: a robustly optimized BERT pretraining approach. arXiv preprint arXiv:1907.11692 (2019). · arXiv
  12. Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017). · arXiv
  13. Rodrigo Nogueira, Wei Yang, Jimmy Lin, and Kyunghyun Cho. 2019. Document expansion by query prediction. arXiv preprint arXiv:1904.08375 (2019). · arXiv
  14. Lee Xiong, Chenyan Xiong, Ye Li, Kwok-Fung Tang, Jialin Liu, Paul Bennett, Junaid Ahmed, and Arnold Overwijk. 2020. Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval. arXiv preprint arXiv:2007.00808 (2020). · arXiv
  15. Yang You, Jing Li, Sashank Reddi, Jonathan Hseu, Sanjiv Kumar, Srinadh Bhojanapalli, Xiaodan Song, James Demmel, Kurt Keutzer, and Cho-Jui Hsieh. 2019. Large batch optimization for deep learning: Training bert in 76 minutes. arXiv preprint arXiv:1904.00962 (2019). · arXiv
  16. Jingtao Zhan, Jiaxin Mao, Yiqun Liu, Min Zhang, and Shaoping Ma. 2020. RepBERT: Contextualized Text Embeddings for First-Stage Retrieval. arXiv preprint arXiv:2006.15498 (2020). · arXiv

Cited by

Sorted by citations

Related

Authors
Zhan, J., Mao, J., Liu, Y., Guo, J., Zhang, M., & Ma, S.
Affiliations
  • Tsinghua University (CN)
  • Renmin University of China (CN)
  • Institute of Computing Technology (CN)
ORCID
Venue
Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval
Publisher
ACM
Published
2021-07-11
Pages
1503-1512
DOI
10.1145/3404835.3462880
OpenAlex
W3155895380
Semantic Scholar
7b577ba0e4230b2ac58d297b3d2cfc3d2f1aaace
Cited by
248
References
18
Citations by year
2026: 19 · 2025: 35 · 2024: 62 · 2023: 76 · 2022: 45 · 2021: 11
FWCI
40.15
Type
conference paper
Language
en
Open access
Licence
www.acm.org/publications/policies/copyright_policy#Background (vor)
Locations
  • International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Funders
  • Natural Science Foundation of China (61732008,61532011)
  • National Key Research and Development Program of China (2018YFC0831700)
  • Tsinghua University Guoqiang Research Institute
  • Beijing Academy of Artificial Intelligence
Topics
  • Physical Sciences › Computer Science › Information Systems › Information Retrieval and Search Behavior
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
Keywords
dense retrieval, hard negative sampling, training strategies, ranking optimization, training stability, stars, ADORE, word matching, vocabulary mismatch, ad hoc retrieval

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review