{"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/the-cot-collection-improving-zero-shot-and","title":"The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning","arxiv_id":"2305.14045","date":"2023-05-23","proceeding":null,"authors":["Seungone Kim","Se June Joo","Doyoung Kim","Joel Jang","Seonghyeon Ye","Jamin Shin","Minjoon Seo"],"abstract":"Language models (LMs) with less than 100B parameters are known to perform poorly on chain-of-thought (CoT) reasoning in contrast to large LMs when solving unseen tasks. In this work, we aim to equip smaller LMs with the step-by-step reasoning capability by instruction tuning with CoT rationales. In order to achieve this goal, we first introduce a new instruction-tuning dataset called the CoT Collection, which augments the existing Flan Collection (including only 9 CoT tasks) with additional 1.84 million rationales across 1,060 tasks. We show that CoT fine-tuning Flan-T5 (3B & 11B) with CoT Collection enables smaller LMs to have better CoT capabilities on unseen tasks. On the BIG-Bench-Hard (BBH) benchmark, we report an average improvement of +4.34% (Flan-T5 3B) and +2.60% (Flan-T5 11B), in terms of zero-shot task accuracy. Furthermore, we show that instruction tuning with CoT Collection allows LMs to possess stronger few-shot learning capabilities on 4 domain-specific tasks, resulting in an improvement of +2.24% (Flan-T5 3B) and +2.37% (Flan-T5 11B), even outperforming ChatGPT utilizing demonstrations until the max length by a +13.98% margin. Our code, the CoT Collection data, and model checkpoints are publicly available.","url_abs":"https://arxiv.org/abs/2305.14045v2","url_pdf":"https://arxiv.org/pdf/2305.14045v2.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":"the-cot-collection-improving-zero-shot-and","repo_url":"https://github.com/kaist-lklab/cot-collection","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"the-cot-collection-improving-zero-shot-and","repo_url":"https://github.com/kaistai/cot-collection","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"common-sense-reasoning","task_name":"Common Sense Reasoning"},{"task_slug":"common-sense-reasoning-zero-shot","task_name":"Common Sense Reasoning (Zero-Shot)"},{"task_slug":"coreference-resolution","task_name":"Coreference Resolution"},{"task_slug":"few-shot-learning","task_name":"Few-Shot Learning"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"natural-language-inference-zero-shot","task_name":"Natural Language Inference (Zero-Shot)"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"sentence-completion","task_name":"Sentence Completion"},{"task_slug":"word-sense-disambiguation","task_name":"Word Sense Disambiguation"}],"methods":[{"method_slug":"flan-t5","method_name":"Flan-T5"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/on-big-bench-hyperbaton","task":"","dataset":"BIG-bench (Hyperbaton)","model":"CoT-T5 11B","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"65.2"},"uses_additional_data":false},{"leaderboard":"/sota/on-big-bench-navigate","task":"","dataset":"BIG-bench (Navigate)","model":"CoT-T5 11B","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"60"},"uses_additional_data":false},{"leaderboard":"/sota/on-big-bench-ruin-names","task":"","dataset":"BIG-bench (Ruin Names)","model":"CoT-T5 11B","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"42.8"},"uses_additional_data":false},{"leaderboard":"/sota/on-big-bench-snarks","task":"","dataset":"BIG-bench (SNARKS)","model":"CoT-T5 11B","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"67.7"},"uses_additional_data":false},{"leaderboard":"/sota/on-big-bench-hard","task":"","dataset":"Big-bench Hard","model":"CoT-T5 11B","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"48"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-winogrande","task":"Common Sense Reasoning","dataset":"WinoGrande","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":57,"of":77,"metrics":{"Accuracy":"57.5"},"uses_additional_data":false},{"leaderboard":"/sota/coreference-resolution-on-winograd-schema","task":"Coreference Resolution","dataset":"Winograd Schema Challenge","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":41,"of":82,"metrics":{"Accuracy":"66"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-learning-on-casehold","task":"Few-Shot Learning","dataset":"CaseHOLD","model":"CoT-T5-11B (1024 Shot)","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"68.3"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-learning-on-mednli","task":"Few-Shot Learning","dataset":"MedNLI","model":"CoT-T5-11B (1024 Shot)","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"78.02"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-learning-on-pubmedqa","task":"Few-Shot Learning","dataset":"PubMedQA","model":"CoT-T5-11B (1024 Shot)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy":"73.42"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-anli-test","task":"Natural Language Inference","dataset":"ANLI test","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":11,"of":25,"metrics":{"A1":"41.7","A2":"37.2","A3":"41.9"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-rte","task":"Natural Language Inference","dataset":"RTE","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":34,"of":90,"metrics":{"Accuracy":"80.8%"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-copa","task":"Question Answering","dataset":"COPA","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":16,"of":60,"metrics":{"Accuracy":"90.9"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-pubmedqa","task":"Question Answering","dataset":"PubMedQA","model":"CoT-T5-11B (1024 Shot)","rank_in_archive_order":21,"of":30,"metrics":{"Accuracy":"73.42"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-storycloze","task":"Question Answering","dataset":"StoryCloze","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":4,"of":23,"metrics":{"Accuracy":"94.5"},"uses_additional_data":false},{"leaderboard":"/sota/sentence-completion-on-hellaswag","task":"Sentence Completion","dataset":"HellaSwag","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":75,"of":89,"metrics":{"Accuracy":"41.1"},"uses_additional_data":false},{"leaderboard":"/sota/word-sense-disambiguation-on-words-in-context","task":"Word Sense Disambiguation","dataset":"Words in Context","model":"T0-3B (CoT fine-tuned)","rank_in_archive_order":21,"of":37,"metrics":{"Accuracy":"56.7"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2305.14045","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}