{"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/finetuned-language-models-are-zero-shot","title":"Finetuned Language Models Are Zero-Shot Learners","arxiv_id":"2109.01652","date":"2021-09-03","proceeding":"ICLR 2022 4","authors":["Jason Wei","Maarten Bosma","Vincent Y. Zhao","Kelvin Guu","Adams Wei Yu","Brian Lester","Nan Du","Andrew M. Dai","Quoc V. Le"],"abstract":"This paper explores a simple method for improving the zero-shot learning abilities of language models. We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks. We take a 137B parameter pretrained language model and instruction-tune it on over 60 NLP tasks verbalized via natural language instruction templates. We evaluate this instruction-tuned model, which we call FLAN, on unseen task types. FLAN substantially improves the performance of its unmodified counterpart and surpasses zero-shot 175B GPT-3 on 20 of 25 tasks that we evaluate. FLAN even outperforms few-shot GPT-3 by a large margin on ANLI, RTE, BoolQ, AI2-ARC, OpenbookQA, and StoryCloze. Ablation studies reveal that number of finetuning datasets, model scale, and natural language instructions are key to the success of instruction tuning.","url_abs":"https://arxiv.org/abs/2109.01652v5","url_pdf":"https://arxiv.org/pdf/2109.01652v5.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":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/google-research/flan","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/bigcode-project/starcoder","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/bigscience-workshop/promptsource","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/hiyouga/llama-efficient-tuning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/hojjat-mokhtarabadi/promptsource","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/openbiolink/promptsource","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/ukplab/arxiv2025-inherent-limits-plms","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"finetuned-language-models-are-zero-shot","repo_url":"https://github.com/MS-P3/code6/tree/main/finetune","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"arc","task_name":"ARC"},{"task_slug":"common-sense-reasoning","task_name":"Common Sense Reasoning"},{"task_slug":"coreference-resolution","task_name":"Coreference Resolution"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"rte","task_name":"RTE"},{"task_slug":"sentence-completion","task_name":"Sentence Completion"},{"task_slug":"sentiment-analysis","task_name":"Sentiment Analysis"},{"task_slug":"zero-shot-learning","task_name":"Zero-Shot Learning"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"attention-dropout","method_name":"Attention Dropout"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"cosine-annealing","method_name":"Cosine Annealing"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gpt-3","method_name":"GPT-3"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"linear-warmup-with-cosine-annealing","method_name":"Linear Warmup With Cosine Annealing"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/common-sense-reasoning-on-arc-challenge","task":"Common Sense Reasoning","dataset":"ARC (Challenge)","model":"FLAN 137B (few-shot, k=13)","rank_in_archive_order":23,"of":54,"metrics":{"Accuracy":"63.8"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-challenge","task":"Common Sense Reasoning","dataset":"ARC (Challenge)","model":"FLAN 137B (zero-shot)","rank_in_archive_order":24,"of":54,"metrics":{"Accuracy":"63.1"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-easy","task":"Common Sense Reasoning","dataset":"ARC (Easy)","model":"FLAN 137B (few-shot, k=14)","rank_in_archive_order":13,"of":47,"metrics":{"Accuracy":"80.7"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-easy","task":"Common Sense Reasoning","dataset":"ARC (Easy)","model":"FLAN 137B (0-shot)","rank_in_archive_order":17,"of":47,"metrics":{"Accuracy":"79.6"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-record","task":"Common Sense Reasoning","dataset":"ReCoRD","model":"FLAN 137B (prompt-tuned)","rank_in_archive_order":13,"of":45,"metrics":{"EM":"85.1"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-record","task":"Common Sense Reasoning","dataset":"ReCoRD","model":"FLAN 137B (zero-shot)","rank_in_archive_order":23,"of":45,"metrics":{"EM":"72.5"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-winogrande","task":"Common Sense Reasoning","dataset":"WinoGrande","model":"FLAN 137B (few-shot, k=16)","rank_in_archive_order":34,"of":77,"metrics":{"Accuracy":"72.8"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-winogrande","task":"Common Sense Reasoning","dataset":"WinoGrande","model":"FLAN 137B (0-shot)","rank_in_archive_order":36,"of":77,"metrics":{"Accuracy":"71.2"},"uses_additional_data":false},{"leaderboard":"/sota/coreference-resolution-on-winograd-schema","task":"Coreference Resolution","dataset":"Winograd Schema Challenge","model":"FLAN 137B (prompt-tuned)","rank_in_archive_order":15,"of":82,"metrics":{"Accuracy":"86.5"},"uses_additional_data":false},{"leaderboard":"/sota/coreference-resolution-on-winograd-schema","task":"Coreference Resolution","dataset":"Winograd Schema Challenge","model":"FLAN 137B (zero-shot)","rank_in_archive_order":20,"of":82,"metrics":{"Accuracy":"80.8"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-french","task":"Machine Translation","dataset":"WMT2014 English-French","model":"FLAN 137B (zero-shot)","rank_in_archive_order":49,"of":57,"metrics":{"BLEU score":"33.9"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-french","task":"Machine Translation","dataset":"WMT2014 English-French","model":"FLAN 137B (few-shot, k=9)","rank_in_archive_order":50,"of":57,"metrics":{"BLEU score":"33.8"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-french-english","task":"Machine Translation","dataset":"WMT2014 French-English","model":"FLAN 137B (few-shot, k=9)","rank_in_archive_order":1,"of":3,"metrics":{"BLEU score":"37.9"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-french-english","task":"Machine Translation","dataset":"WMT2014 French-English","model":"FLAN 137B (zero-shot)","rank_in_archive_order":2,"of":3,"metrics":{"BLEU score":"35.9"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-german","task":"Machine Translation","dataset":"WMT2016 English-German","model":"FLAN 137B (zero-shot)","rank_in_archive_order":5,"of":12,"metrics":{"BLEU score":"27.0"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-german","task":"Machine Translation","dataset":"WMT2016 English-German","model":"FLAN 137B (few-shot, k=11)","rank_in_archive_order":7,"of":12,"metrics":{"BLEU score":"26.1"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-1","task":"Machine Translation","dataset":"WMT2016 English-Romanian","model":"FLAN 137B (few-shot, k=9)","rank_in_archive_order":19,"of":21,"metrics":{"BLEU score":"20.5"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-1","task":"Machine Translation","dataset":"WMT2016 English-Romanian","model":"FLAN 137B (zero-shot)","rank_in_archive_order":20,"of":21,"metrics":{"BLEU score":"18.9"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-german-english","task":"Machine Translation","dataset":"WMT2016 German-English","model":"FLAN 137B (few-shot, k=11)","rank_in_archive_order":1,"of":8,"metrics":{"BLEU score":"40.7"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-german-english","task":"Machine Translation","dataset":"WMT2016 German-English","model":"FLAN 137B (zero-shot)","rank_in_archive_order":2,"of":8,"metrics":{"BLEU score":"38.9"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-romanian","task":"Machine Translation","dataset":"WMT2016 Romanian-English","model":"FLAN 137B (few-shot, k=9)","rank_in_archive_order":2,"of":21,"metrics":{"BLEU score":"38.1"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-romanian","task":"Machine Translation","dataset":"WMT2016 Romanian-English","model":"FLAN 137B (zero-shot)","rank_in_archive_order":3,"of":21,"metrics":{"BLEU score":"37.3"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-rte","task":"Natural Language Inference","dataset":"RTE","model":"FLAN 137B (prompt-tuned)","rank_in_archive_order":13,"of":90,"metrics":{"Accuracy":"91.7%"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-rte","task":"Natural Language Inference","dataset":"RTE","model":"FLAN 137B (8-shot)","rank_in_archive_order":29,"of":90,"metrics":{"Accuracy":"84.5%"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-rte","task":"Natural Language Inference","dataset":"RTE","model":"FLAN 137B (0-shot)","rank_in_archive_order":30,"of":90,"metrics":{"Accuracy":"84.1%"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-wnli","task":"Natural Language Inference","dataset":"WNLI","model":"FLAN 137B (zero-shot)","rank_in_archive_order":14,"of":23,"metrics":{"Accuracy":"74.6"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-wnli","task":"Natural Language Inference","dataset":"WNLI","model":"FLAN 137B (few-shot, k=4)","rank_in_archive_order":17,"of":23,"metrics":{"Accuracy":"70.4"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-boolq","task":"Question Answering","dataset":"BoolQ","model":"FLAN 137B (prompt-tuned)","rank_in_archive_order":14,"of":65,"metrics":{"Accuracy":"86.3"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-boolq","task":"Question Answering","dataset":"BoolQ","model":"FLAN 137B (4-shot)","rank_in_archive_order":19,"of":65,"metrics":{"Accuracy":"84.6"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-boolq","task":"Question Answering","dataset":"BoolQ","model":"FLAN 137B (0-shot)","rank_in_archive_order":24,"of":65,"metrics":{"Accuracy":"82.9"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-copa","task":"Question Answering","dataset":"COPA","model":"FLAN 137B (prompt-tuned)","rank_in_archive_order":10,"of":60,"metrics":{"Accuracy":"94"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-copa","task":"Question Answering","dataset":"COPA","model":"FLAN 137B (zero-shot)","rank_in_archive_order":13,"of":60,"metrics":{"Accuracy":"91"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-copa","task":"Question Answering","dataset":"COPA","model":"FLAN 137B (few-shot, k=16)","rank_in_archive_order":22,"of":60,"metrics":{"Accuracy":"87"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-multirc","task":"Question Answering","dataset":"MultiRC","model":"FLAN 137B (prompt-tuned)","rank_in_archive_order":11,"of":30,"metrics":{"F1":"83.4"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-multirc","task":"Question Answering","dataset":"MultiRC","model":"FLAN 137B (zero-shot)","rank_in_archive_order":12,"of":30,"metrics":{"F1":"77.5"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-multirc","task":"Question Answering","dataset":"MultiRC","model":"FLAN 137B (1-shot)","rank_in_archive_order":14,"of":30,"metrics":{"F1":"72.1"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-naturalqa","task":"Question Answering","dataset":"NaturalQA","model":"FLAN 137B (zero-shot)","rank_in_archive_order":2,"of":4,"metrics":{"EM":"20.7"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-obqa","task":"Question Answering","dataset":"OBQA","model":"FLAN 137B (zero-shot)","rank_in_archive_order":1,"of":9,"metrics":{"Accuracy":"78.4"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-obqa","task":"Question Answering","dataset":"OBQA","model":"FLAN 137B (few-shot, k=16)","rank_in_archive_order":2,"of":9,"metrics":{"Accuracy":"78.2"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-piqa","task":"Question Answering","dataset":"PIQA","model":"FLAN 137B (few-shot, k=10)","rank_in_archive_order":27,"of":67,"metrics":{"Accuracy":"81.7"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-piqa","task":"Question Answering","dataset":"PIQA","model":"FLAN 137B (0-shot)","rank_in_archive_order":31,"of":67,"metrics":{"Accuracy":"80.5"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-storycloze","task":"Question Answering","dataset":"StoryCloze","model":"FLAN 137B (few-shot, k=10)","rank_in_archive_order":3,"of":23,"metrics":{"Accuracy":"94.7"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-storycloze","task":"Question Answering","dataset":"StoryCloze","model":"FLAN 137B (zero-shot)","rank_in_archive_order":6,"of":23,"metrics":{"Accuracy":"93.4"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-triviaqa","task":"Question Answering","dataset":"TriviaQA","model":"FLAN 137B (zero-shot)","rank_in_archive_order":42,"of":56,"metrics":{"EM":"56.7"},"uses_additional_data":false},{"leaderboard":"/sota/sentence-completion-on-hellaswag","task":"Sentence Completion","dataset":"HellaSwag","model":"FLAN 137B (3-shot)","rank_in_archive_order":60,"of":89,"metrics":{"Accuracy":"59.2"},"uses_additional_data":false},{"leaderboard":"/sota/sentence-completion-on-hellaswag","task":"Sentence Completion","dataset":"HellaSwag","model":"FLAN 137B (0-shot)","rank_in_archive_order":62,"of":89,"metrics":{"Accuracy":"56.7"},"uses_additional_data":false},{"leaderboard":"/sota/sentiment-analysis-on-imdb","task":"Sentiment Analysis","dataset":"IMDb","model":"FLAN 137B (few-shot, k=2)","rank_in_archive_order":18,"of":49,"metrics":{"Accuracy":"95"},"uses_additional_data":true},{"leaderboard":"/sota/sentiment-analysis-on-imdb","task":"Sentiment Analysis","dataset":"IMDb","model":"FLAN 137B (zero-shot)","rank_in_archive_order":22,"of":49,"metrics":{"Accuracy":"94.3"},"uses_additional_data":true}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2109.01652","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2109.01652"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bigscience-workshop/promptsource","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/hiyouga/llama-efficient-tuning","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/hojjat-mokhtarabadi/promptsource","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bigcode-project/starcoder","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/openbiolink/promptsource","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/MS-P3/code6/tree/main/finetune","reach":null},{"provenance":"deterministic:regex_extraction","url":"https://github.com/google-research/flan","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ukplab/arxiv2025-inherent-limits-plms","reach":null}],"summary":{"unverified":1},"by_repo_kind":{"listed":{"samples":1,"ran":0,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"76f691da79652f18","entry":"preprocess_flan_dataset","repo":"ukplab/arxiv2025-inherent-limits-plms","repo_kind":"listed","path":"data_utils.py","file_url":"https://github.com/ukplab/arxiv2025-inherent-limits-plms/blob/HEAD/data_utils.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"76f691da79652f18"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}