{"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/unifying-language-learning-paradigms","title":"UL2: Unifying Language Learning Paradigms","arxiv_id":"2205.05131","date":"2022-05-10","proceeding":null,"authors":["Yi Tay","Mostafa Dehghani","Vinh Q. Tran","Xavier Garcia","Jason Wei","Xuezhi Wang","Hyung Won Chung","Siamak Shakeri","Dara Bahri","Tal Schuster","Huaixiu Steven Zheng","Denny Zhou","Neil Houlsby","Donald Metzler"],"abstract":"Existing pre-trained models are generally geared towards a particular class of problems. To date, there seems to be still no consensus on what the right architecture and pre-training setup should be. This paper presents a unified framework for pre-training models that are universally effective across datasets and setups. We begin by disentangling architectural archetypes with pre-training objectives -- two concepts that are commonly conflated. Next, we present a generalized & unified perspective for self-supervision in NLP and show how different pre-training objectives can be cast as one another and how interpolating between different objectives can be effective. We then propose Mixture-of-Denoisers (MoD), a pre-training objective that combines diverse pre-training paradigms together. We furthermore introduce a notion of mode switching, wherein downstream fine-tuning is associated with specific pre-training schemes. We conduct extensive ablative experiments to compare multiple pre-training objectives and find that our method pushes the Pareto-frontier by outperforming T5 & GPT-like models across multiple diverse setups. By scaling our model up to 20B parameters, we achieve SOTA performance on 50 well-established supervised finetuning based NLP tasks. Our model also achieve strong results at in-context learning, outperforming 175B GPT-3 on zero-shot SuperGLUE and tripling the performance of T5-XXL on one-shot summarization. On 0-shot MMLU, UL2 20B outperforms T0 and T5 models. UL2 20B also works well with chain-of-thought prompting and reasoning, making it an appealing choice for research into reasoning at a small to medium scale of 20B parameters. Finally, we apply FLAN instruction tuning to the UL2 20B model, achieving MMLU and Big-Bench scores competitive to FLAN-PaLM 62B. We release Flax-based T5X checkpoints for the UL2 20B & Flan-UL2 20B.","url_abs":"https://arxiv.org/abs/2205.05131v3","url_pdf":"https://arxiv.org/pdf/2205.05131v3.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":"unifying-language-learning-paradigms","repo_url":"https://github.com/google-research/google-research","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"unifying-language-learning-paradigms","repo_url":"https://github.com/opennlg/openba-v2","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"arithmetic-reasoning","task_name":"Arithmetic Reasoning"},{"task_slug":"common-sense-reasoning","task_name":"Common Sense Reasoning"},{"task_slug":"coreference-resolution","task_name":"Coreference Resolution"},{"task_slug":"in-context-learning","task_name":"In-Context Learning"},{"task_slug":"information-retrieval","task_name":"Information Retrieval"},{"task_slug":"long-range-modeling","task_name":"Long-range modeling"},{"task_slug":"mmlu","task_name":"MMLU"},{"task_slug":"multi-task-language-understanding","task_name":"Multi-task Language Understanding"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"text-classification","task_name":"Text Classification"},{"task_slug":"text-generation","task_name":"Text Generation"},{"task_slug":"word-sense-disambiguation","task_name":"Word Sense Disambiguation"}],"methods":[{"method_slug":"adafactor","method_name":"Adafactor"},{"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":"glu","method_name":"Gated Linear Unit"},{"method_slug":"inverse-square-root-schedule","method_name":"Inverse Square Root Schedule"},{"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":"sentencepiece","method_name":"SentencePiece"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"t5","method_name":"T5"},{"method_slug":"ul2","method_name":"UL2"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[],"methods_introduced":[{"slug":"ul2","name":"UL2","full_name":"UL2"}],"results":[{"leaderboard":"/sota/arithmetic-reasoning-on-gsm8k","task":"Arithmetic Reasoning","dataset":"GSM8K","model":"UL2 20B (chain-of-thought)","rank_in_archive_order":162,"of":164,"metrics":{"Accuracy":"4.4","Parameters (Billion)":"20"},"uses_additional_data":false},{"leaderboard":"/sota/arithmetic-reasoning-on-gsm8k","task":"Arithmetic Reasoning","dataset":"GSM8K","model":"UL2 20B (0-shot)","rank_in_archive_order":164,"of":164,"metrics":{"Accuracy":"4.1","Parameters (Billion)":"20"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-challenge","task":"Common Sense Reasoning","dataset":"ARC (Challenge)","model":"UL2 20B (chain-of-thought + self-consistency)","rank_in_archive_order":37,"of":54,"metrics":{"Accuracy":"49.5"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-challenge","task":"Common Sense Reasoning","dataset":"ARC (Challenge)","model":"UL2 20B (chain-of-thought)","rank_in_archive_order":45,"of":54,"metrics":{"Accuracy":"42.9"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-challenge","task":"Common Sense Reasoning","dataset":"ARC (Challenge)","model":"UL2 20B (zero-shot)","rank_in_archive_order":53,"of":54,"metrics":{"Accuracy":"29.8"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-easy","task":"Common Sense Reasoning","dataset":"ARC (Easy)","model":"UL2 20B (chain-of-thought + self-consistency)","rank_in_archive_order":34,"of":47,"metrics":{"Accuracy":"69.8"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-easy","task":"Common Sense Reasoning","dataset":"ARC (Easy)","model":"UL2 20B (chain-of-thought)","rank_in_archive_order":44,"of":47,"metrics":{"Accuracy":"38.4"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-arc-easy","task":"Common Sense Reasoning","dataset":"ARC (Easy)","model":"UL2 20B (0-shot)","rank_in_archive_order":46,"of":47,"metrics":{"Accuracy":"32.2"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-commonsenseqa","task":"Common Sense Reasoning","dataset":"CommonsenseQA","model":"UL2 20B (chain-of-thought + self-consistency)","rank_in_archive_order":32,"of":38,"metrics":{"Accuracy":"55.7"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-commonsenseqa","task":"Common Sense Reasoning","dataset":"CommonsenseQA","model":"UL2 20B (chain-of-thought)","rank_in_archive_order":34,"of":38,"metrics":{"Accuracy":"51.4"},"uses_additional_data":false},{"leaderboard":"/sota/common-sense-reasoning-on-commonsenseqa","task":"Common Sense Reasoning","dataset":"CommonsenseQA","model":"UL2 20B (zero-shot)","rank_in_archive_order":36,"of":38,"metrics":{"Accuracy":"34.2"},"uses_additional_data":false},{"leaderboard":"/sota/coreference-resolution-on-winograd-schema","task":"Coreference Resolution","dataset":"Winograd Schema Challenge","model":"UL2 20B (fine-tuned)","rank_in_archive_order":3,"of":82,"metrics":{"Accuracy":"98.1"},"uses_additional_data":false},{"leaderboard":"/sota/coreference-resolution-on-winograd-schema","task":"Coreference Resolution","dataset":"Winograd Schema Challenge","model":"UL2 20B (0-shot)","rank_in_archive_order":23,"of":82,"metrics":{"Accuracy":"79.9"},"uses_additional_data":false},{"leaderboard":"/sota/long-range-modeling-on-scrolls","task":"Long-range modeling","dataset":"SCROLLS","model":"UL2","rank_in_archive_order":7,"of":13,"metrics":{"Avg.":"37.87","GovRep":"53.6 / 26.1 / 28.8","Nrtv":"24.2","QALT EM-T/H":"45.8 / 40.7","QMSum":"31.1 / 8.5 / 20.4","Qspr":"37.6","SumScr":"32.9 / 7.8 / 19.4"},"uses_additional_data":false},{"leaderboard":"/sota/long-range-modeling-on-scrolls","task":"Long-range modeling","dataset":"SCROLLS","model":"UL2 20B","rank_in_archive_order":13,"of":13,"metrics":{"CNLI":"88.7"},"uses_additional_data":false},{"leaderboard":"/sota/multi-task-language-understanding-on-mmlu","task":"Multi-task Language Understanding","dataset":"MML","model":"UL2 20B (5-shot)","rank_in_archive_order":34,"of":44,"metrics":{"Average (%)":"39.2"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-rte","task":"Natural Language Inference","dataset":"RTE","model":"UL2 20B (fine-tuned)","rank_in_archive_order":11,"of":90,"metrics":{"Accuracy":"92.1%"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-rte","task":"Natural Language Inference","dataset":"RTE","model":"UL2 20B (0-shot)","rank_in_archive_order":71,"of":90,"metrics":{"Accuracy":"60.7%"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-boolq","task":"Question Answering","dataset":"BoolQ","model":"UL2 20B (fine-tuned)","rank_in_archive_order":7,"of":65,"metrics":{"Accuracy":"90.8"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-boolq","task":"Question Answering","dataset":"BoolQ","model":"UL2 20B (0-shot)","rank_in_archive_order":50,"of":65,"metrics":{"Accuracy":"63.1"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-copa","task":"Question Answering","dataset":"COPA","model":"UL2 20B (fine-tuned)","rank_in_archive_order":4,"of":60,"metrics":{"Accuracy":"99"},"uses_additional_data":false},{"leaderboard":"/sota/question-answering-on-copa","task":"Question Answering","dataset":"COPA","model":"UL2 20B (0-shot)","rank_in_archive_order":29,"of":60,"metrics":{"Accuracy":"85"},"uses_additional_data":false},{"leaderboard":"/sota/word-sense-disambiguation-on-words-in-context","task":"Word Sense Disambiguation","dataset":"Words in Context","model":"UL2 20B (fine-tuned)","rank_in_archive_order":6,"of":37,"metrics":{"Accuracy":"77.3"},"uses_additional_data":false},{"leaderboard":"/sota/word-sense-disambiguation-on-words-in-context","task":"Word Sense Disambiguation","dataset":"Words in Context","model":"UL2 20B (0-shot)","rank_in_archive_order":35,"of":37,"metrics":{"Accuracy":"49.8"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2205.05131","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2205.05131"}},"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/google-research/google-research","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/opennlg/openba-v2","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"unverified":16},"by_repo_kind":{"listed":{"samples":16,"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":"b87d8f19cf535a2a","entry":"add_checkpointing_args","repo":"opennlg/openba-v2","repo_kind":"listed","path":"convert_megatron_to_hf_ckpt.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/convert_megatron_to_hf_ckpt.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":"b87d8f19cf535a2a"}},{"code_sha256_prefix":"33a078e319b0a5aa","entry":"add_megatron_checkpoint_args","repo":"opennlg/openba-v2","repo_kind":"listed","path":"convert_megatron_to_hf_ckpt.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/convert_megatron_to_hf_ckpt.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":"33a078e319b0a5aa"}},{"code_sha256_prefix":"e126c3b654292801","entry":"add_transformers_checkpoint_args","repo":"opennlg/openba-v2","repo_kind":"listed","path":"convert_megatron_to_hf_ckpt.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/convert_megatron_to_hf_ckpt.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":"e126c3b654292801"}},{"code_sha256_prefix":"5fd8991277f9c276","entry":"get_input","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/ARC/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/ARC/template.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":"5fd8991277f9c276"}},{"code_sha256_prefix":"b24c9bcfbcc13329","entry":"make_ABCD_input_0_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/ARC/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/ARC/template.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":"b24c9bcfbcc13329"}},{"code_sha256_prefix":"706a6589174ca080","entry":"make_ABCD_input_0_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/C-Eval/moban.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/C-Eval/moban.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":"706a6589174ca080"}},{"code_sha256_prefix":"9ca4a81f11223f44","entry":"make_ABCD_input_0_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/CMMLU/moban.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/CMMLU/moban.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":"9ca4a81f11223f44"}},{"code_sha256_prefix":"7c7bc3af994e8c59","entry":"make_ABCD_input_25_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/ARC/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/ARC/template.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":"7c7bc3af994e8c59"}},{"code_sha256_prefix":"e2eaf8f3ca6c3150","entry":"make_ppl_input_0_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/C-Eval/moban.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/C-Eval/moban.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":"e2eaf8f3ca6c3150"}},{"code_sha256_prefix":"2f649837afcee3d7","entry":"make_ppl_input_0_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/MMLU/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/MMLU/template.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":"2f649837afcee3d7"}},{"code_sha256_prefix":"e6bb902901596a24","entry":"make_ppl_input_0_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/PIQA/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/PIQA/template.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":"e6bb902901596a24"}},{"code_sha256_prefix":"c10614604fc54cfe","entry":"make_ppl_input_10_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/PIQA/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/PIQA/template.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":"c10614604fc54cfe"}},{"code_sha256_prefix":"259357a18b0eb827","entry":"make_ppl_input_5_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/C-Eval/moban.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/C-Eval/moban.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":"259357a18b0eb827"}},{"code_sha256_prefix":"092eedfde7c54d94","entry":"make_ppl_input_5_shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/MMLU/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/MMLU/template.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":"092eedfde7c54d94"}},{"code_sha256_prefix":"563f98855addfe7a","entry":"make_r_ppl_input_10shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/PIQA/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/PIQA/template.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":"563f98855addfe7a"}},{"code_sha256_prefix":"9dfbe0e29d0f6312","entry":"make_r_ppl_input_5shot","repo":"opennlg/openba-v2","repo_kind":"listed","path":"evaluation/MMLU/template.py","file_url":"https://github.com/opennlg/openba-v2/blob/HEAD/evaluation/MMLU/template.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":"9dfbe0e29d0f6312"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}