Research map / Label-efficient and robust learning / Recognition and long-tail
Learning to Tokenize for Generative Retrieval
Weiwei Sun, Lingyong Yan, Zheng Chen, Shuaiqiang Wang, Haichao Zhu, Pengjie Ren, Zhumin Chen, Dawei Yin, Maarten de Rijke, Zhaochun Ren
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| remove docid re-assignment (Sinkhorn) from diverse clustering (ablation) | GenRet · NQ320K first-step docid distribution | diversity d: 0.77 -> 0.90 | Improved |
| add diverse codebook initialization (constrained k-means) (ablation) | GenRet · NQ320K docid distribution | docid balance | Improved |
| use progressive auto-encoding learning instead of training directly on final docids (ablation) | GenRet · NQ320K unseen test | R@1 | Improved |
| use T5-Base instead of T5-Small (ablation) | GenRet · NQ320K | retrieval performance | Improved |
| replace rule-based docids (BERT-HC) with learned docids | generative retrieval · NQ unseen (zero-shot) | R@1: 1.3 -> 34.2 | Improved |
| replace Prefix-HC docids (NCI) with learned docids | generative retrieval · NQ unseen (zero-shot) | R@1: 15.5 -> 34.2 | Improved |
| replace title docids (GENRE) with learned docids | generative retrieval · MS MARCO | MRR@10: 42.3 -> 58.1 | Improved |
| replace title docids (GENRE) with learned docids | generative retrieval · BEIR average | nDCG@10: 30.0 -> 41.1 | Improved |
| GenRet versus best dense retriever | GenRet · BEIR average | nDCG@10: 41.1 vs 41.8 (Contriever) | No clear effect |
Closest papers by meaning
- Descriptive and Discriminative Document Identifiers for Generative Retrieval (AAAI 2025 · similarity 0.90)
- Generative Retrieval Meets Multi-Graded Relevance (NeurIPS 2024 · similarity 0.90)
- DOGR: Leveraging Document-Oriented Contrastive Learning in Generative Retrieval (AAAI 2025 · similarity 0.89)
- Learning to Rank in Generative Retrieval (AAAI 2024 · similarity 0.88)
- Retrieval is Accurate Generation (ICLR 2024 · similarity 0.87)
- Language Models as Semantic Indexers (ICML 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
