Papers › UL2: Unifying Language Learning Paradigms

UL2: Unifying Language Learning Paradigms

10 May 2022arXiv:2205.05131archive 2025-07-28

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

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.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

No sample from the archive's official repository (google-research/google-research) is linked to this paper; the 16 samples below come from 1 community repository. Syntology harvested code from google-research/google-research for 24 other papers.

google-research/google-research: Syntology's graph links these samples to Efficient Neural Architecture Search via Parameter Sharing (arXiv:1802.03268; that paper's own run record: 18 ran (of which 10 constructed an object rather than computing a result; 12 with no instrument failure: 1 honoured, 0 violated, 11 with no contract checked; 6 where Syntology's instrument failed) · 20 unverified), Milking CowMask for Semi-Supervised Image Classification (arXiv:2003.12022; that paper's own run record: 1 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified; the one sample that ran constructed an object rather than computing a result), Neural Additive Models: Interpretable Machine Learning with Neural Nets (arXiv:2004.13912; that paper's own run record: 26 ran (of which 14 constructed an object rather than computing a result; 26 with no instrument failure: 5 honoured, 0 violated, 21 with no contract checked; 0 where Syntology's instrument failed) · 7 unverified) and 21 more (every one in this page's JSON twin); nothing checks that they implement this paper's method.

Syntology 16 code samples harvested from 1 repository linked to this paper: 15 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 0 honoured, 0 violated, 15 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code; here there are none.

By repository: community (archive-listed): 16 samples from 1 repository, 15 ran (15 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 1 of its repositories that Syntology harvested for other papers:

Levels written on 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

google-research/google-research officialmentioned in papertfSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Efficient Neural Architecture Search via Parameter Sharing (arXiv:1802.03268; that paper's own run record: 18 ran (of which 10 constructed an object rather than computing a result; 12 with no instrument failure: 1 honoured, 0 violated, 11 with no contract checked; 6 where Syntology's instrument failed) · 20 unverified), Milking CowMask for Semi-Supervised Image Classification (arXiv:2003.12022; that paper's own run record: 1 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified; the one sample that ran constructed an object rather than computing a result), Neural Additive Models: Interpretable Machine Learning with Neural Nets (arXiv:2004.13912; that paper's own run record: 26 ran (of which 14 constructed an object rather than computing a result; 26 with no instrument failure: 5 honoured, 0 violated, 21 with no contract checked; 0 where Syntology's instrument failed) · 7 unverified) and 21 more (every one in this page's JSON twin); nothing checks that they implement this paper's method. report
opennlg/openba-v2 mentioned on GitHubpytorchApache-2.0Syntology: 16 samples, 15 ran (0 instrument-failed) · 1 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

16 samples harvested: 15 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 0 honoured, 0 violated, 15 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

15ran
1unverified

Licence: 0 of the 16 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from opennlg/openba-v2. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: 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. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

opennlg/openba-v2 community (archive-listed) 16 samples, 15 ran (0 instrument-failed) · 1 unverified
add_checkpointing_args convert_megatron_to_hf_ckpt.py ran Apache-2.0 (permissive) · b87d8f19cf535a2a · report fixture: signature parser:ArgumentParser/0/? · returned ArgumentParser · level recorded 2026-09-24
add_transformers_checkpoint_args convert_megatron_to_hf_ckpt.py ran Apache-2.0 (permissive) · e126c3b654292801 · report fixture: signature parser:MockParser/0/? · returned MockParser · level recorded 2026-09-24
get_input in 20 papers evaluation/ARC/template.py ran Apache-2.0 (permissive) · 5fd8991277f9c276 · report fixture: signature demo:dict/0/? · returned str · level recorded 2026-09-24
make_ABCD_input_0_shot evaluation/ARC/template.py ran Apache-2.0 (permissive) · b24c9bcfbcc13329 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ABCD_input_0_shot evaluation/C-Eval/moban.py ran Apache-2.0 (permissive) · 706a6589174ca080 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ABCD_input_0_shot evaluation/CMMLU/moban.py ran Apache-2.0 (permissive) · 9ca4a81f11223f44 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ABCD_input_25_shot evaluation/ARC/template.py ran Apache-2.0 (permissive) · 7c7bc3af994e8c59 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ppl_input_0_shot evaluation/C-Eval/moban.py ran Apache-2.0 (permissive) · e2eaf8f3ca6c3150 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ppl_input_0_shot evaluation/MMLU/template.py ran Apache-2.0 (permissive) · 2f649837afcee3d7 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ppl_input_0_shot evaluation/PIQA/template.py ran Apache-2.0 (permissive) · e6bb902901596a24 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ppl_input_10_shot evaluation/PIQA/template.py ran Apache-2.0 (permissive) · c10614604fc54cfe · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ppl_input_5_shot evaluation/C-Eval/moban.py ran Apache-2.0 (permissive) · 259357a18b0eb827 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_ppl_input_5_shot evaluation/MMLU/template.py ran Apache-2.0 (permissive) · 092eedfde7c54d94 · report fixture: signature subject:str/0/str, data:dict/0/? · returned tuple · level recorded 2026-09-24
make_r_ppl_input_10shot evaluation/PIQA/template.py ran Apache-2.0 (permissive) · 563f98855addfe7a · report fixture: signature subject:str/0/str, data:dict/0/? · returned list · level recorded 2026-09-24
make_r_ppl_input_5shot evaluation/MMLU/template.py ran Apache-2.0 (permissive) · 9dfbe0e29d0f6312 · report fixture: signature subject:str/0/str, data:dict/0/? · returned list · level recorded 2026-09-24
add_megatron_checkpoint_args convert_megatron_to_hf_ckpt.py unverified Apache-2.0 (permissive) · 33a078e319b0a5aa · report

Tasks

Arithmetic ReasoningCommon Sense ReasoningCoreference ResolutionIn-Context LearningInformation RetrievalLong-range modelingMMLUMulti-task Language UnderstandingNatural Language InferenceQuestion AnsweringRetrievalText ClassificationText GenerationWord Sense Disambiguation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Arithmetic Reasoning GSM8K UL2 20B (chain-of-thought) Accuracy 4.4 #162 of 164 Archive leaderboard report
Arithmetic Reasoning GSM8K UL2 20B (chain-of-thought) Parameters (Billion) 20 #162 of 164 Archive leaderboard report
Arithmetic Reasoning GSM8K UL2 20B (0-shot) Accuracy 4.1 #164 of 164 Archive leaderboard report
Arithmetic Reasoning GSM8K UL2 20B (0-shot) Parameters (Billion) 20 #164 of 164 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) UL2 20B (chain-of-thought + self-consistency) Accuracy 49.5 #37 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) UL2 20B (chain-of-thought) Accuracy 42.9 #45 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) UL2 20B (zero-shot) Accuracy 29.8 #53 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Easy) UL2 20B (chain-of-thought + self-consistency) Accuracy 69.8 #34 of 47 Archive leaderboard report
Common Sense Reasoning ARC (Easy) UL2 20B (chain-of-thought) Accuracy 38.4 #44 of 47 Archive leaderboard report
Common Sense Reasoning ARC (Easy) UL2 20B (0-shot) Accuracy 32.2 #46 of 47 Archive leaderboard report
Common Sense Reasoning CommonsenseQA UL2 20B (chain-of-thought + self-consistency) Accuracy 55.7 #32 of 38 Archive leaderboard report
Common Sense Reasoning CommonsenseQA UL2 20B (chain-of-thought) Accuracy 51.4 #34 of 38 Archive leaderboard report
Common Sense Reasoning CommonsenseQA UL2 20B (zero-shot) Accuracy 34.2 #36 of 38 Archive leaderboard report
Coreference Resolution Winograd Schema Challenge UL2 20B (fine-tuned) Accuracy 98.1 #3 of 82 Archive leaderboard report
Coreference Resolution Winograd Schema Challenge UL2 20B (0-shot) Accuracy 79.9 #23 of 82 Archive leaderboard report
Long-range modeling SCROLLS UL2 Avg. 37.87 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 GovRep 53.6 / 26.1 / 28.8 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 Nrtv 24.2 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 QALT EM-T/H 45.8 / 40.7 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 QMSum 31.1 / 8.5 / 20.4 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 Qspr 37.6 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 SumScr 32.9 / 7.8 / 19.4 #7 of 13 Archive leaderboard report
Long-range modeling SCROLLS UL2 20B CNLI 88.7 #13 of 13 Archive leaderboard report
Multi-task Language Understanding MML UL2 20B (5-shot) Average (%) 39.2 #34 of 44 Archive leaderboard report
Natural Language Inference RTE UL2 20B (fine-tuned) Accuracy 92.1% #11 of 90 Archive leaderboard report
Natural Language Inference RTE UL2 20B (0-shot) Accuracy 60.7% #71 of 90 Archive leaderboard report
Question Answering BoolQ UL2 20B (fine-tuned) Accuracy 90.8 #7 of 65 Archive leaderboard report
Question Answering BoolQ UL2 20B (0-shot) Accuracy 63.1 #50 of 65 Archive leaderboard report
Question Answering COPA UL2 20B (fine-tuned) Accuracy 99 #4 of 60 Archive leaderboard report
Question Answering COPA UL2 20B (0-shot) Accuracy 85 #29 of 60 Archive leaderboard report
Word Sense Disambiguation Words in Context UL2 20B (fine-tuned) Accuracy 77.3 #6 of 37 Archive leaderboard report
Word Sense Disambiguation Words in Context UL2 20B (0-shot) Accuracy 49.8 #35 of 37 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Introduced by this paper: UL2

AdafactorAdamAttentionAttention DropoutBPECosine AnnealingDense ConnectionsDropoutGPT-3Gated Linear UnitInverse Square Root ScheduleLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionResidual ConnectionSentencePieceSoftmaxT5UL2Weight Decay

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections