{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/large-dual-encoders-are-generalizable","title":"Large Dual Encoders Are Generalizable Retrievers","arxiv_id":"2112.07899","date":"2021-12-15","proceeding":null,"authors":["Jianmo Ni","Chen Qu","Jing Lu","Zhuyun Dai","Gustavo Hernández Ábrego","Ji Ma","Vincent Y. Zhao","Yi Luan","Keith B. Hall","Ming-Wei Chang","Yinfei Yang"],"abstract":"It has been shown that dual encoders trained on one domain often fail to generalize to other domains for retrieval tasks. One widespread belief is that the bottleneck layer of a dual encoder, where the final score is simply a dot-product between a query vector and a passage vector, is too limited to make dual encoders an effective retrieval model for out-of-domain generalization. In this paper, we challenge this belief by scaling up the size of the dual encoder model {\\em while keeping the bottleneck embedding size fixed.} With multi-stage training, surprisingly, scaling up the model size brings significant improvement on a variety of retrieval tasks, especially for out-of-domain generalization. Experimental results show that our dual encoders, \\textbf{G}eneralizable \\textbf{T}5-based dense \\textbf{R}etrievers (GTR), outperform %ColBERT~\\cite{khattab2020colbert} and existing sparse and dense retrievers on the BEIR dataset~\\cite{thakur2021beir} significantly. Most surprisingly, our ablation study finds that GTR is very data efficient, as it only needs 10\\% of MS Marco supervised data to achieve the best out-of-domain performance. All the GTR models are released at https://tfhub.dev/google/collections/gtr/1.","url_abs":"https://arxiv.org/abs/2112.07899v1","url_pdf":"https://arxiv.org/pdf/2112.07899v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"large-dual-encoders-are-generalizable","repo_url":"https://github.com/google-research/t5x_retrieval","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"jax","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"large-dual-encoders-are-generalizable","repo_url":"https://github.com/openmatch/coco-dr","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"domain-generalization","task_name":"Domain Generalization"},{"task_slug":"passage-retrieval","task_name":"Passage Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"sentence-retrieval","task_name":"Sentence Retrieval"},{"task_slug":"zero-shot-text-search","task_name":"Zero-shot Text Search"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/passage-retrieval-on-peerqa","task":"Passage Retrieval","dataset":"PeerQA","model":"GTR-XL","rank_in_archive_order":7,"of":8,"metrics":{"MRR":"0.4142","Recall@10":"0.6122"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2112.07899","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}