{"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/multitask-prompted-training-enables-zero-shot-1","title":"Multitask Prompted Training Enables Zero-Shot Task Generalization","arxiv_id":"2110.08207","date":"2021-10-15","proceeding":"ICLR 2022 4","authors":["Victor Sanh","Albert Webson","Colin Raffel","Stephen H. Bach","Lintang Sutawika","Zaid Alyafeai","Antoine Chaffin","Arnaud Stiegler","Teven Le Scao","Arun Raja","Manan Dey","M Saiful Bari","Canwen Xu","Urmish Thakker","Shanya Sharma Sharma","Eliza Szczechla","Taewoon Kim","Gunjan Chhablani","Nihal Nayak","Debajyoti Datta","Jonathan Chang","Mike Tian-Jian Jiang","Han Wang","Matteo Manica","Sheng Shen","Zheng Xin Yong","Harshit Pandey","Rachel Bawden","Thomas Wang","Trishala Neeraj","Jos Rozen","Abheesht Sharma","Andrea Santilli","Thibault Fevry","Jason Alan Fries","Ryan Teehan","Tali Bers","Stella Biderman","Leo Gao","Thomas Wolf","Alexander M. Rush"],"abstract":"Large language models have recently been shown to attain reasonable zero-shot generalization on a diverse set of tasks (Brown et al., 2020). It has been hypothesized that this is a consequence of implicit multitask learning in language models' pretraining (Radford et al., 2019). Can zero-shot generalization instead be directly induced by explicit multitask learning? To test this question at scale, we develop a system for easily mapping any natural language tasks into a human-readable prompted form. We convert a large set of supervised datasets, each with multiple prompts with diverse wording. These prompted datasets allow for benchmarking the ability of a model to perform completely held-out tasks. We fine-tune a pretrained encoder-decoder model (Raffel et al., 2020; Lester et al., 2021) on this multitask mixture covering a wide variety of tasks. The model attains strong zero-shot performance on several standard datasets, often outperforming models up to 16x its size. Further, our approach attains strong performance on a subset of tasks from the BIG-bench benchmark, outperforming models up to 6x its size. All trained models are available at https://github.com/bigscience-workshop/t-zero and all prompts are available at https://github.com/bigscience-workshop/promptsource.","url_abs":"https://arxiv.org/abs/2110.08207v3","url_pdf":"https://arxiv.org/pdf/2110.08207v3.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":"multitask-prompted-training-enables-zero-shot-1","repo_url":"https://github.com/bigscience-workshop/promptsource","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"multitask-prompted-training-enables-zero-shot-1","repo_url":"https://github.com/bigscience-workshop/t-zero","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"multitask-prompted-training-enables-zero-shot-1","repo_url":"https://github.com/DevSinghSachan/art","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"multitask-prompted-training-enables-zero-shot-1","repo_url":"https://github.com/facebookresearch/metaicl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"multitask-prompted-training-enables-zero-shot-1","repo_url":"https://github.com/gonglinyuan/metro_t0","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"multitask-prompted-training-enables-zero-shot-1","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":"multitask-prompted-training-enables-zero-shot-1","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":"multitask-prompted-training-enables-zero-shot-1","repo_url":"https://github.com/ukplab/arxiv2022-context-injection-stance","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"benchmarking","task_name":"Benchmarking"},{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"prompt-engineering","task_name":"Prompt Engineering"},{"task_slug":"zero-shot-generalization","task_name":"Zero-shot Generalization"}],"methods":[{"method_slug":"test","method_name":"Test"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2110.08207","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2110.08207"}},"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":"deterministic:regex_extraction","url":"https://github.com/bigscience-workshop/promptsource","reach":null},{"provenance":"deterministic:regex_extraction","url":"https://github.com/bigscience-workshop/t-zero","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/DevSinghSachan/art","reach":null},{"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/openbiolink/promptsource","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/ukplab/arxiv2022-context-injection-stance","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/gonglinyuan/metro_t0","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/facebookresearch/metaicl","reach":null}],"summary":{"ran_violates":1,"ran":5,"ran_draft_wrong":1,"ran_fixture":1,"unverified":7},"by_repo_kind":{"official":{"samples":6,"ran":6,"repositories":2},"listed":{"samples":9,"ran":2,"repositories":2}},"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":"bb35fcd18dd842fb","entry":"choice","repo":"bigscience-workshop/promptsource","repo_kind":"official","path":"promptsource/templates.py","file_url":"https://github.com/bigscience-workshop/promptsource/blob/HEAD/promptsource/templates.py","link_basis":"first_harvest_node","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"bb35fcd18dd842fb"}},{"code_sha256_prefix":"b7af70f17bfb5e07","entry":"feature_to_spec","repo":"bigscience-workshop/t-zero","repo_kind":"official","path":"t0/seqio_tasks/utils.py","file_url":"https://github.com/bigscience-workshop/t-zero/blob/HEAD/t0/seqio_tasks/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"b7af70f17bfb5e07"}},{"code_sha256_prefix":"de62b3eab371bdf6","entry":"hf_dataset_to_tf_dataset","repo":"bigscience-workshop/t-zero","repo_kind":"official","path":"t0/seqio_tasks/utils.py","file_url":"https://github.com/bigscience-workshop/t-zero/blob/HEAD/t0/seqio_tasks/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"de62b3eab371bdf6"}},{"code_sha256_prefix":"406e3e3a9886da66","entry":"highlight","repo":"bigscience-workshop/promptsource","repo_kind":"official","path":"promptsource/templates.py","file_url":"https://github.com/bigscience-workshop/promptsource/blob/HEAD/promptsource/templates.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"406e3e3a9886da66"}},{"code_sha256_prefix":"10c4b8cfda850b4c","entry":"most_frequent","repo":"bigscience-workshop/promptsource","repo_kind":"official","path":"promptsource/templates.py","file_url":"https://github.com/bigscience-workshop/promptsource/blob/HEAD/promptsource/templates.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"10c4b8cfda850b4c"}},{"code_sha256_prefix":"0b1cb631e6e315b0","entry":"removeHyphen","repo":"hojjat-mokhtarabadi/promptsource","repo_kind":"listed","path":"promptsource/utils.py","file_url":"https://github.com/hojjat-mokhtarabadi/promptsource/blob/HEAD/promptsource/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"0b1cb631e6e315b0"}},{"code_sha256_prefix":"a77010010d56e618","entry":"renameDatasetColumn","repo":"hojjat-mokhtarabadi/promptsource","repo_kind":"listed","path":"promptsource/utils.py","file_url":"https://github.com/hojjat-mokhtarabadi/promptsource/blob/HEAD/promptsource/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a77010010d56e618"}},{"code_sha256_prefix":"11b64b0a85dfb964","entry":"strip_whitespace","repo":"bigscience-workshop/t-zero","repo_kind":"official","path":"t0/seqio_tasks/tasks.py","file_url":"https://github.com/bigscience-workshop/t-zero/blob/HEAD/t0/seqio_tasks/tasks.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"11b64b0a85dfb964"}},{"code_sha256_prefix":"9b27d71357f65d09","entry":"fake_query","repo":"ukplab/arxiv2022-context-injection-stance","repo_kind":"listed","path":"src/retrieve/retrieve_t0pp_context.py","file_url":"https://github.com/ukplab/arxiv2022-context-injection-stance/blob/HEAD/src/retrieve/retrieve_t0pp_context.py","link_basis":"harvester_set","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":"9b27d71357f65d09"}},{"code_sha256_prefix":"aa86817b5ef44488","entry":"get_attribution_stats","repo":"ukplab/arxiv2022-context-injection-stance","repo_kind":"listed","path":"src/analysis/analyse_attributions.py","file_url":"https://github.com/ukplab/arxiv2022-context-injection-stance/blob/HEAD/src/analysis/analyse_attributions.py","link_basis":"harvester_set","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":"aa86817b5ef44488"}},{"code_sha256_prefix":"e14d51a44fec9477","entry":"get_dataset_builder","repo":"hojjat-mokhtarabadi/promptsource","repo_kind":"listed","path":"promptsource/utils.py","file_url":"https://github.com/hojjat-mokhtarabadi/promptsource/blob/HEAD/promptsource/utils.py","link_basis":"harvester_set","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":"e14d51a44fec9477"}},{"code_sha256_prefix":"35a244de1ccdbabd","entry":"get_divergence","repo":"ukplab/arxiv2022-context-injection-stance","repo_kind":"listed","path":"src/analysis/analyse_topics.py","file_url":"https://github.com/ukplab/arxiv2022-context-injection-stance/blob/HEAD/src/analysis/analyse_topics.py","link_basis":"harvester_set","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":"35a244de1ccdbabd"}},{"code_sha256_prefix":"dd07b1fdb8bc71a0","entry":"get_indomain_splits","repo":"ukplab/arxiv2022-context-injection-stance","repo_kind":"listed","path":"src/analysis/count_dataset_splits.py","file_url":"https://github.com/ukplab/arxiv2022-context-injection-stance/blob/HEAD/src/analysis/count_dataset_splits.py","link_basis":"harvester_set","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":"dd07b1fdb8bc71a0"}},{"code_sha256_prefix":"f296f02f0a3d87ac","entry":"jensen_shannon_divergence","repo":"ukplab/arxiv2022-context-injection-stance","repo_kind":"listed","path":"src/analysis/analyse_topics.py","file_url":"https://github.com/ukplab/arxiv2022-context-injection-stance/blob/HEAD/src/analysis/analyse_topics.py","link_basis":"harvester_set","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":"f296f02f0a3d87ac"}},{"code_sha256_prefix":"279ad86d58c62ffb","entry":"query","repo":"ukplab/arxiv2022-context-injection-stance","repo_kind":"listed","path":"src/retrieve/retrieve_t0pp_context.py","file_url":"https://github.com/ukplab/arxiv2022-context-injection-stance/blob/HEAD/src/retrieve/retrieve_t0pp_context.py","link_basis":"harvester_set","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":"279ad86d58c62ffb"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}