Elamur Research⌕
☰
home / papers / 2026 / QueStER: Query Specification for Generative Keyword-Based Retrieval

QueStER: Query Specification for Generative Keyword-Based Retrieval

Paper · 2026 · Arthur Satouf, Yuxuan Zong, Habiboulaye Amadou Boubacar et al.

Summary

Generative retrieval stores relevance knowledge inside model weights. This makes it fragile on unfamiliar domains and costly to scale. Neural first-stage retrievers have their own problem: they need large indices that must be rebuilt every time the model is retrained. QueStER takes a different route. A small LLM (Qwen3-4B) does not produce document identifiers. Instead, it writes a keyword "search specification" for each query, and a plain BM25 engine over an ordinary inverted index runs that query.

The rewriting policy is trained with GRPO, using no KL penalty. The reward is SoftNDCG, an expected-nDCG measure built on SoftRank. It is computed from graded relevance labels produced by a cross-encoder. Training queries come from MS MARCO, and evaluation covers 15 BEIR datasets.

Compared with BM25, QueStER gains about 4 nDCG@10 points in-domain and about 5 points out-of-domain. It roughly matches prompt-based rewriting methods that use much larger LLMs and is competitive with neural IR out of domain. The approach matters because it keeps lexical indexing cheap and its queries readable while still delivering much of the benefit of learned models.

Key points

  • Gains over BM25 in nDCG@10: +4.0 in-domain and +5.3 out-of-domain, at roughly 28 ms per query for the BM25 stage.
  • Index cost motivates the design: a dense FP16 index of MS MARCO (8.8M passages) needs about 13 GiB, while a BM25 index needs about 0.67 GiB.
  • Cross-encoder-derived graded labels beat sparse qrels by about 1 nDCG@10 point.
    • The best SoftRank smoothing is ν=0.5, with a 45.5 average over 15 BEIR datasets.
    • β=0 works best; β=0.05 drops the average to 40.9.
  • Warm-starting with SFT did not beat RL alone. The SFT data was 50k MS MARCO rewrites generated with Qwen3-32B.
  • The authors chose the prompt after testing more than 50 candidates on a 1k-query MS MARCO subset.
  • Appending the original query at inference adds only about +0.1 nDCG@10.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksInformation Systems · 1,437 works

🧭 Topics 1Research subjects and languages the work is about.Information retrieval & search · 473 works

Cite

Cite as Arthur Satouf, Yuxuan Zong, Habiboulaye Amadou Boubacar, Pablo Piantanida and Benjamin Piwowarski. 2026. QueStER: Query Specification for Generative Keyword-Based Retrieval. Findings of the Association for Computational Linguistics: EACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-eacl.312

@inproceedings{satouf2026quester,
  title = {{QueStER: Query Specification for Generative Keyword-Based Retrieval}},
  author = {Arthur Satouf, Yuxuan Zong, Habiboulaye Amadou Boubacar, Pablo Piantanida and Benjamin Piwowarski},
  booktitle = {Findings of the Association for Computational Linguistics: EACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-eacl.312},
  url = {https://doi.org/10.18653/v1/2026.findings-eacl.312}
}

References

59 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Nasreen Abdul-Jaleel, James Allan, W. Bruce Croft, Fernando Diaz, Leah Larkey, Xiaoyan Li, Mark D. Smucker, and Courtney Wade. 2004. Umass at trec 2004: Novelty and hard. In Text REtrieval Conference (TREC). NIST. find
  2. Marwah Alaofi, Luke Gallagher, Mark Sanderson, Falk Scholer, and Paul Thomas. 2023. Can generative llms create query variants for test collections? an exploratory study. In ACM Conference on Research and Development in Information Retrieval (SIGIR). find
  3. Payal Bajaj, Daniel Campos, Nick Craswell, Li Deng, Jianfeng Gao, Xiaodong Liu, Rangan Majumder, Andrew McNamara, Bhaskar Mitra, Tri Nguyen, and 1 others. 2016. Ms marco: A human generated machine reading comprehension dataset. arXiv preprint arXiv:1611.09268.
  4. Soyuj Basnet, Jerry Gou, Antonio Mallia, and Torsten Suel. 2024. Deeperimpact: Optimizing sparse learned index structures. Preprint, arXiv:2405.17093. — [arXiv:2405.17093](https://arxiv.org/abs/2405.17093) find
  5. Michele Bevilacqua, Giuseppe Ottaviano, Patrick Lewis, Scott Yih, Sebastian Riedel, and Fabio Petroni. 2022. Autoregressive search engines: Generating substrings as document identifiers. Advances in Neural Information Processing Systems, 35:31668–31683. find
  6. Claudio Carpineto and Giovanni Romano. 2012. A survey of automatic query expansion in information retrieval. Acm Computing Surveys (CSUR), 44(1):1– 50. find
  7. Jiangui Chen, Ruqing Zhang, Jiafeng Guo, Yixing Fan, and Xueqi Cheng. 2022a. Gere: Generative evidence retrieval for fact verification. In ACM Conference on Research and Development in Information Retrieval (SIGIR), pages 2184–2189. find
  8. Jiangui Chen, Ruqing Zhang, Jiafeng Guo, Yiqun Liu, Yixing Fan, and Xueqi Cheng. 2022b. Corpusbrain: Pre-train a generative retrieval model for knowledge-intensive language tasks. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pages 191–200. find
  9. Nicola De Cao, Gautier Izacard, Sebastian Riedel, and Fabio Petroni. 2020. Autoregressive entity retrieval. arXiv preprint arXiv:2010.00904. find
  10. Thibault Formal, Carlos Lassance, Benjamin Piwowarski, and Stéphane Clinchant. 2021. Splade v2: Sparse lexical and expansion model for information retrieval. arXiv preprint arXiv:2109.10086.
  11. Luyu Gao, Xueguang Ma, Jimmy Lin, and Jamie Callan. 2023. Precise zero-shot dense retrieval without relevance labels. In Proceedings of the Annual Meeting of the Association for Computational Linguistics, pages 1762–1777.
  12. Adrien Grand, Robert Muir, Jim Ferenczi, and Jimmy Lin. 2020. From MAXSCORE to Block-Max Wand: The Story of How Lucene Significantly Improved Query Evaluation Performance. In Advances in Information Retrieval, Lecture Notes in Computer Science, pages 20–27, Cham. Springer International Publishing. find
  13. Michael Granitzer, Stefan Voigt, Noor Afshan Fathima, Martin Golasowski, Christian Guetl, Tobias Hecking, Gijs Hendriksen, Djoerd Hiemstra, Jan Martinovič, Jelena Mitrović, Izidor Mlakar, Stavros Moiras, Alexander Nussbaumer, Per Öster, Martin Potthast, Marjana Senčar Srdič, Sharikadze Megi, Kat… find
  14. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  15. Prashansa Gupta and Sean MacAvaney. 2022. On survivorship bias in ms marco. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’22, page 2214–2219. ACM. find
  16. Rolf Jagerman, Honglei Zhuang, Zhen Qin, Xuanhui Wang, and Michael Bendersky. 2023. Query expansion by prompting large language models. arXiv preprint library, abs/2305.03653. — [arXiv:2305.03653](https://arxiv.org/abs/2305.03653) find
  17. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6769–6781. Associati…
  18. Victor Lavrenko and W. Bruce Croft. 2001. Relevance-based language models. In ACM Conference on Research and Development in Information Retrieval (SIGIR), pages 120–127. ACM. find
  19. Sunkyung Lee, Minjin Choi, and Jongwuk Lee. 2023. Glen: Generative retrieval via lexical index learning. arXiv preprint arXiv:2311.03057. find
  20. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, and 1 others. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in neural information processing systems, 33:9459–…
  21. Minghan Li, Honglei Zhuang, Kai Hui, Zhen Qin, Jimmy Lin, Rolf Jagerman, Xuanhui Wang, and Michael Bendersky. 2024. Can query expansion improve generalization of strong cross-encoder rankers? In ACM Conference on Research and Development in Information Retrieval (SIGIR), pages 2321–2326. find
  22. Yongqi Li, Nan Yang, Liang Wang, Furu Wei, and Wenjie Li. 2023. Multiview identifiers enhanced generative retrieval. arXiv preprint arXiv:2305.16675. find
  23. Xinyu Ma, Yixing Gong, Pengjie He, Heng Zhao, and Nan Duan. 2023. Query rewriting for retrieval-augmented large language models. In emnlp. find
  24. Sebastian Hofstätter, Sheng-Chieh Lin, Jheng-Hong Yang, Jimmy Lin, and Allan Hanbury. 2021. Efficiently Teaching an Effective Dense Retriever with Balanced Topic Aware Sampling. arXiv:2104.06967 [cs]. ZSCC: 0000000 arXiv: 2104.06967.
  25. Xueguang Ma, Liang Wang, Nan Yang, Furu Wei, and Jimmy Lin. 2024. Finetuning llama for multi stage text retrieval. In ACM Conference on Research and Development in Information Retrieval (SIGIR), pages 3164–3168. find
  26. Sheryl Hsu, Omar Khattab, Chelsea Finn, and Archit Sharma. 2024. Grounding by trying: Llms with reinforcement learning-enhanced retrieval. arXiv preprint arXiv:2410.23214. — [arXiv:2410.23214](https://arxiv.org/abs/2410.23214) find
  27. Iain Mackie, Shubham Chatterjee, and Jeffrey Dalton. 2023. Generative and pseudo-relevant feedback for sparse, dense and learned sparse retrieval. arXiv preprint library, abs/2305.07477. — [arXiv:2305.07477](https://arxiv.org/abs/2305.07477) find
  28. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3.
  29. Shengyu Mao, Yong Jiang, Boli Chen, Xiao Li, Peng Wang, Xinyu Wang, Pengjun Xie, Fei Huang, Huajun Chen, and Ningyu Zhang. 2024. Rafe: Ranking feedback improves query rewriting for rag. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  30. Jingcheng Hu, Yinmin Zhang, Qi Han, Daxin Jiang, Xiangyu Zhang, and Heung-Yeung Shum. 2025. Open-reasoner-zero: An open source approach to scaling up reinforcement learning on the base model. arXiv preprint arXiv:2503.24290.
  31. Mandar Mitra, Amit Singhal, and Chris Buckley. 1998. Improving automatic query expansion. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’98, page 206–214, New York, NY, USA. Association for Computing Machinery. find
  32. Rodrigo Nogueira and Kyunghyun Cho. 2017. Task-oriented query reformulation with reinforcement learning. Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  33. Rodrigo Nogueira and Kyunghyun Cho. 2019. Passage re-ranking with bert. arXiv preprint arXiv:1901.04085.
  34. Aaron van den Oord, Yazhe Li, and Oriol Vinyals. 2018. Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748.
  35. Seongwan Park, Taeklim Kim, and Youngjoong Ko. 2025. Decoding dense embeddings: Sparse autoencoders for interpreting and discretizing dense retrieval. arXiv preprint arXiv:2506.00041. find
  36. Wenjun Peng, Guiyang Li, Yue Jiang, Zilong Wang, Dan Ou, Xiaoyi Zeng, Derong Xu, Tong Xu, and Enhong Chen. 2024. Large language model based long-tail query rewriting in taobao search. In Companion Proceedings of the ACM Web Conference 2024, pages 20–28. find
  37. Ronak Pradeep, Kai Hui, Jai Gupta, Adam D Lelkes, Honglei Zhuang, Jimmy Lin, Donald Metzler, and Vinh Q Tran. 2023. How does generative retrieval scale to millions of passages? arXiv preprint arXiv:2305.11841. find
  38. Stephen Robertson, Steve Walker, Susan Jones, Micheline Hancock-Beaulieu, and Mike Gatford. 1994. Okapi at trec-3. In "", pages 0–.
  39. Keshav Santhanam, Omar Khattab, Jon Saad-Falcon, Christopher Potts, and Matei Zaharia. 2021. Colbertv2: Effective and efficient retrieval via lightweight late interaction. arXiv preprint arXiv:2112.01488.
  40. Arthur Satouf, Gabriel Ben Zenou, Benjamin Piwowarski, Habiboulaye Amadou Boubacar, and Pablo Piantanida. 2025. Rational retrieval acts: Leveraging pragmatic reasoning to improve sparse retrieval. Preprint, arXiv:2505.03676. — [arXiv:2505.03676](https://arxiv.org/abs/2505.03676) find
  41. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, and 1 others. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300.
  42. Tao Shen, Guodong Long, Xiubo Geng, Chongyang Tao, Tianze Zhou, and Daxin Jiang. 2024. Retrieval-augmented retrieval: Large language models are strong zero-shot retriever. Proceedings of the Annual Meeting of the Association for Computational Linguistics.
  43. Weiwei Sun, Lingyong Yan, Zheng Chen, Shuaiqiang Wang, Haichao Zhu, Pengjie Ren, Zhumin Chen, Dawei Yin, Maarten Rijke, and Zhaochun Ren. 2023. Learning to tokenize for generative retrieval. Advances in Neural Information Processing Systems, 36:46345–46361. find
  44. Yubao Tang, Ruqing Zhang, Jiafeng Guo, Jiangui Chen, Zuowei Zhu, Shuaiqiang Wang, Dawei Yin, and Xueqi Cheng. 2023. Semantic-enhanced differentiable search index inspired by learning strategies. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 4904–4913. find
  45. Yi Tay, Vinh Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Gupta, and 1 others. 2022. Transformer memory as a differentiable search index. Advances in Neural Information Processing Systems, 35:21831–21843.
  46. Michael Taylor, John Guiver, Stephen Robertson, and Tom Minka. 2008. Softrank: optimizing non-smooth rank metrics. In Proceedings of the 2008 International Conference on Web Search and Data Mining, WSDM ’08, page 77–86, New York, NY, USA. Association for Computing Machinery. find
  47. Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, and 1 others. 2025. Gemma 3 technical report. arXiv preprint arXiv:2503.19786. find
  48. Nandan Thakur, Nils Reimers, Andreas Rücklé, Abhishek Srivastava, and Iryna Gurevych. 2021. Beir: A heterogenous benchmark for zero-shot evaluation of information retrieval models. arXiv preprint arXiv:2104.08663.
  49. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In neurips.
  50. Liang Wang, Nan Yang, and Furu Wei. 2023. Query2doc: Query expansion with large language models. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP).
  51. Yujing Wang, Yingyan Hou, Haonan Wang, Ziming Miao, Shibin Wu, Qi Chen, Yuqing Xia, Chengmin Chi, Guoshuai Zhao, Zheng Liu, and 1 others. 2022. A neural corpus indexer for document retrieval. Advances in Neural Information Processing Systems, 35:25600–25614. find
  52. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  53. Tianchi Yang, Minghui Song, Zihan Zhang, Haizhen Huang, Weiwei Deng, Feng Sun, and Qi Zhang. 2023. find
  54. Auto search indexer for end-to-end document retrieval. arXiv preprint arXiv:2310.12455. find
  55. Sijia Yao, Pengcheng Huang, Zhenghao Liu, Yu Gu, Yukun Yan, Shi Yu, and Ge Yu. 2025. Llm-qe: Improving query expansion by aligning large language models with ranking preferences. arXiv preprint arXiv:2502.17057. — [arXiv:2502.17057](https://arxiv.org/abs/2502.17057) find
  56. Fanghua Ye, Meng Fang, Shenghui Li, and Emine Yilmaz. 2023. Enhancing conversational search: Large language model-aided informative query rewriting. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  57. Le Zhang, Yihong Wu, Qian Yang, and Jian-Yun Nie. 2024a. Exploring the best practices of query expansion with large language models. Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  58. Peitian Zhang, Zheng Liu, Yujia Zhou, Zhicheng Dou, Fangchao Liu, and Zhao Cao. 2024b. Generative retrieval via term set generation. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 458–468. find
  59. Yujia Zhou, Jing Yao, Zhicheng Dou, Ledell Wu, Peitian Zhang, and Ji-Rong Wen. 2022. Ultron: An ultimate retriever on corpus with a model-based indexer. arXiv preprint arXiv:2208.09257. — [arXiv:2208.09257](https://arxiv.org/abs/2208.09257) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/satouf-2026-quester-query-specification-generative-keyword/