QueStER: Query Specification for Generative keyword-based Retrieval
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911390079385600 |
|---|---|
| author | Satouf, Arthur Zong, Yuxuan Amadou-Boubacar, Habiboulaye Piantanida, Pablo Piwowarski, Benjamin |
| author_facet | Satouf, Arthur Zong, Yuxuan Amadou-Boubacar, Habiboulaye Piantanida, Pablo Piwowarski, Benjamin |
| contents | Generative retrieval (GR) differs from the traditional index-then-retrieve pipeline by storing relevance in model parameters and generating retrieval cues directly from the query, but it can be brittle out of domain and expensive to scale. We introduce QueStER (QUEry SpecificaTion for gEnerative Keyword-Based Retrieval), which bridges GR and query reformulation by learning to generate explicit keyword-based search specifications. Given a user query, a lightweight LLM produces a keyword query that is executed by a standard retriever (BM25), combining the generalization benefits of generative query rewriting with the efficiency and scalability of lexical indexing. We train the rewriting policy with reinforcement learning techniques. Across in- and out-of-domain evaluations, QueStER consistently improves over BM25 and is competitive with neural IR baselines, while maintaining strong efficiency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_05301 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | QueStER: Query Specification for Generative keyword-based Retrieval Satouf, Arthur Zong, Yuxuan Amadou-Boubacar, Habiboulaye Piantanida, Pablo Piwowarski, Benjamin Information Retrieval Computation and Language Machine Learning 68P20, 68T50 H.3 Generative retrieval (GR) differs from the traditional index-then-retrieve pipeline by storing relevance in model parameters and generating retrieval cues directly from the query, but it can be brittle out of domain and expensive to scale. We introduce QueStER (QUEry SpecificaTion for gEnerative Keyword-Based Retrieval), which bridges GR and query reformulation by learning to generate explicit keyword-based search specifications. Given a user query, a lightweight LLM produces a keyword query that is executed by a standard retriever (BM25), combining the generalization benefits of generative query rewriting with the efficiency and scalability of lexical indexing. We train the rewriting policy with reinforcement learning techniques. Across in- and out-of-domain evaluations, QueStER consistently improves over BM25 and is competitive with neural IR baselines, while maintaining strong efficiency. |
| title | QueStER: Query Specification for Generative keyword-based Retrieval |
| topic | Information Retrieval Computation and Language Machine Learning 68P20, 68T50 H.3 |
| url | https://arxiv.org/abs/2511.05301 |