{"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/no-parameter-left-behind-how-distillation-and","title":"No Parameter Left Behind: How Distillation and Model Size Affect Zero-Shot Retrieval","arxiv_id":"2206.02873","date":"2022-06-06","proceeding":null,"authors":["Guilherme Moraes Rosa","Luiz Bonifacio","Vitor Jeronymo","Hugo Abonizio","Marzieh Fadaee","Roberto Lotufo","Rodrigo Nogueira"],"abstract":"Recent work has shown that small distilled language models are strong competitors to models that are orders of magnitude larger and slower in a wide range of information retrieval tasks. This has made distilled and dense models, due to latency constraints, the go-to choice for deployment in real-world retrieval applications. In this work, we question this practice by showing that the number of parameters and early query-document interaction play a significant role in the generalization ability of retrieval models. Our experiments show that increasing model size results in marginal gains on in-domain test sets, but much larger gains in new domains never seen during fine-tuning. Furthermore, we show that rerankers largely outperform dense ones of similar size in several tasks. Our largest reranker reaches the state of the art in 12 of the 18 datasets of the Benchmark-IR (BEIR) and surpasses the previous state of the art by 3 average points. Finally, we confirm that in-domain effectiveness is not a good indicator of zero-shot effectiveness. Code is available at https://github.com/guilhermemr04/scaling-zero-shot-retrieval.git","url_abs":"https://arxiv.org/abs/2206.02873v5","url_pdf":"https://arxiv.org/pdf/2206.02873v5.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":"no-parameter-left-behind-how-distillation-and","repo_url":"https://github.com/guilhermemr04/scaling-zero-shot-retrieval","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"argument-retrieval","task_name":"Argument Retrieval"},{"task_slug":"biomedical-information-retrieval","task_name":"Biomedical Information Retrieval"},{"task_slug":"citation-prediction","task_name":"Citation Prediction"},{"task_slug":"duplicate-question-retrieval","task_name":"Duplicate-Question Retrieval"},{"task_slug":"entity-retrieval","task_name":"Entity Retrieval"},{"task_slug":"fact-checking","task_name":"Fact Checking"},{"task_slug":"information-retrieval","task_name":"Information Retrieval"},{"task_slug":"news-retrieval","task_name":"News Retrieval"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"tweet-retrieval","task_name":"Tweet Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/biomedical-information-retrieval-on-bioasq-1","task":"Biomedical Information Retrieval","dataset":"BioASQ (BEIR)","model":"monoT5-3B","rank_in_archive_order":1,"of":6,"metrics":{"nDCG@10":"0.579"},"uses_additional_data":false},{"leaderboard":"/sota/biomedical-information-retrieval-on-nfcorpus-1","task":"Biomedical Information Retrieval","dataset":"NFCorpus (BEIR)","model":"monoT5-3B","rank_in_archive_order":1,"of":7,"metrics":{"nDCG@10":"0.383"},"uses_additional_data":false},{"leaderboard":"/sota/biomedical-information-retrieval-on-trec-1","task":"Biomedical Information Retrieval","dataset":"TREC-COVID (BEIR)","model":"monoT5-3B","rank_in_archive_order":2,"of":6,"metrics":{"nDCG@10":"0.795"},"uses_additional_data":false},{"leaderboard":"/sota/fact-checking-on-climate-fever-beir","task":"Fact Checking","dataset":"CLIMATE-FEVER (BEIR)","model":"monoT5-3B","rank_in_archive_order":2,"of":4,"metrics":{"nDCG@10":"0.280"},"uses_additional_data":false},{"leaderboard":"/sota/fact-checking-on-fever-beir","task":"Fact Checking","dataset":"FEVER (BEIR)","model":"monoT5-3B","rank_in_archive_order":1,"of":4,"metrics":{"nDCG@10":"0.849"},"uses_additional_data":false},{"leaderboard":"/sota/fact-checking-on-scifact-beir","task":"Fact Checking","dataset":"SciFact (BEIR)","model":"monoT5-3B","rank_in_archive_order":1,"of":5,"metrics":{"nDCG@10":"0.777"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-fiqa-2018-beir","task":"Question Answering","dataset":"FiQA-2018 (BEIR)","model":"monoT5-3B","rank_in_archive_order":1,"of":4,"metrics":{"nDCG@10":"0.513"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-hotpotqa-beir","task":"Question Answering","dataset":"HotpotQA (BEIR)","model":"monoT5-3B","rank_in_archive_order":1,"of":4,"metrics":{"nDCG@10":"0.759"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-nq-beir","task":"Question Answering","dataset":"NQ (BEIR)","model":"monoT5-3B","rank_in_archive_order":2,"of":6,"metrics":{"nDCG@10":"0.633"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2206.02873","atlas_url":"https://app.syntology.ai/?focus=2206.02873","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}