Papers › RoBERTa: A Robustly Optimized BERT Pretraining Approach
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov
Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code.
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):
get_harvested_code_for_paper(arxiv_id="1907.11692")get_code(code_sha256="…")with a sample'scode_sha256from the list belowget_citation_path(paper_1="1907.11692", paper_2="…")with another paper's arXiv id or titleget_concepts_for_paper(arxiv_id="1907.11692")
Code
No sample from the archive's official repository (pytorch/fairseq) is linked to this paper; the 48 samples below come from 16 community repositories. Syntology harvested code from pytorch/fairseq for 3 other papers.
pytorch/fairseq: Syntology's graph links these samples to LEWIS: Levenshtein Editing for Unsupervised Text Style Transfer (arXiv:2105.08206; 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) · 0 unverified; the one sample that ran constructed an object rather than computing a result), Cascaded Head-colliding Attention (arXiv:2105.14850; that paper's own run record: 9 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 1 honoured, 1 violated, 2 with no contract checked; 5 where Syntology's instrument failed) · 1 unverified) and Modeling Dual Read/Write Paths for Simultaneous Machine Translation (arXiv:2203.09163; that paper's own run record: 4 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 2 where Syntology's instrument failed) · 5 unverified); nothing checks that they implement this paper's method.
Syntology 48 code samples harvested from 16 repositories linked to this paper: 37 ran (of which 11 constructed an object rather than computing a result; 36 with no instrument failure: 2 honoured, 0 violated, 34 with no contract checked; 1 where Syntology's instrument failed) · 11 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; failures of Syntology's instrument, not of the code.
By repository: community (archive-listed): 48 samples from 16 repositories, 37 ran (36 with no instrument failure; 1 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 14 of its repositories that Syntology harvested for other papers:
- pytorch/fairseq, official repository: harvested for 3 other papers, listed under the repository
- awslabs/mlm-scoring, community (archive-listed): harvested for another paper, listed under the repository
- clovaai/textual-kd-slu, community (archive-listed): harvested for another paper, listed under the repository
- dig-team/hanna-benchmark-asg, community (archive-listed): harvested for another paper, listed under the repository
- huggingface/transformers, community (archive-listed): harvested for 4 other papers, listed under the repository
- knuddj1/op_text, community (archive-listed): harvested for another paper, listed under the repository
- lashoun/hanna-benchmark-asg, community (archive-listed): harvested for another paper, listed under the repository
- mthcom/hscore-dataset-pruning, community (archive-listed): harvested for another paper, listed under the repository
- utterworks/fast-bert, community (archive-listed): harvested for another paper, listed under the repository
- zfj1998/CodeBert-Code2Text, community (archive-listed): harvested for another paper, listed under the repository
- znhy1024/protoco, community (archive-listed): harvested for another paper, listed under the repository
- PaddlePaddle/PaddleNLP, community (archive-listed): harvested for another paper, listed under the repository
- pwc-1/Paper-9, community (archive-listed): harvested for another paper, listed under the repository
- yangyucheng000/University, community (archive-listed): harvested for another paper, listed under the repository
Levels written between 2026-09-22 and 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.
67 repositories listed; official and paper-mentioned ones first.
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
48 samples harvested: 37 ran (of which 11 constructed an object rather than computing a result; 36 with no instrument failure: 2 honoured, 0 violated, 34 with no contract checked; 1 where Syntology's instrument failed) · 11 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.
Licence: 24 of the 48 samples are pointer only, meaning Syntology does not serve that copy's text. For 1 of them, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.
Harvested from 16 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “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”. For 7 samples that ran, the graph records a signature this page does not show: for a class the fixture built without calling, the harness called the object with inputs it made up itself, so the recorded signature is the harness's probe, not the fixture's arguments. “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. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.
ad7f6e57dedf08aa · report
fixture: returned BertAttention · level recorded 2026-09-22 c3af34b4f438eaea · report
fixture: signature inputs_embeds:Tensor/3/float, position_ids:NoneType/0/? · returned Tensor · level recorded 2026-09-24 6e60ffe93c5ec5fa · report
fixture: returned BertIntermediate · level recorded 2026-09-22 f43b4161b7bdba24 · report
fixture: signature inputs_embeds:Tensor/3/float, attention_mask:Tensor/2/float · returned tuple · level recorded 2026-09-24 5e5629cb8e93e004 · report
fixture: signature hidden_states:Tensor/2/float, input_tensor:Tensor/2/float · returned Tensor · level recorded 2026-09-24 4e755eff48bf0752 · report
fixture: returned BertPooler · level recorded 2026-09-22 df8363b5430892a0 · report
fixture: signature hidden_states:Tensor/3/float, attention_mask:NoneType/0/? · returned tuple · level recorded 2026-09-24 fcf3cd282eb03efe · report
fixture: returned BertSelfOutput · level recorded 2026-09-22 00ea7705507db813 · report
f2ab856479b90e8a · report
7b8d60647337c2b1 · report
fixture: returned ClassificationHead · level recorded 2026-09-22 0196b04010ddfa00 · report
fixture: signature input_ids:Tensor/2/int, attention_mask:Tensor/2/float, extra_data:Tensor/2/float, labels:Tensor/1/int · returned SequenceClassifierOutput · level recorded 2026-09-24 92e80bcaa1f41216 · report
da835c893505251b · report
152609ff366dfb9e · report
fixture: signature input_ids:Tensor/2/int, support_class_num:int/0/int, orig_prototypes:NoneType/0/?, orig_embed_class_len:NoneType/0/?, attention_mask:Tensor/2/float, position_ids:NoneType/0/?, head_mask:NoneType/0/?, inputs_embeds:NoneType/0/?, labels:Tensor/2/int · returned tuple · level recorded 2026-09-24 dba0b61c9d62f82f · report
fixture: returned Pooler · level recorded 2026-09-22 e06455e72782e6d0 · report
fixture: signature none_df:DataFrame/0/? · returned DataFrame · level recorded 2026-09-24 d4db5e37c5831f1a · report
fixture: signature row:dict/0/? · returned bool · level recorded 2026-09-24 57b444571a95cb8a · report
fixture: signature row:dict/0/? · returned bool · level recorded 2026-09-24 b3e0c37e1e2b9503 · report
3caf26b311c98658 · report
fixture: signature task_names:str/0/str · returned list · level recorded 2026-09-22 17243795c9ab1ec3 · report
fixture: signature x:Tensor/2/float · returned Tensor · level recorded 2026-09-22 163e38293198fa6a · report
fixture: signature seq:list/1/int · returned list · level recorded 2026-09-24 9f11fd760e57188c · report
fixture: signature probs1:ndarray/2/float, probs2:ndarray/2/float, le:LabelEncoder/0/? · returned list · level recorded 2026-09-24 census: 4 implementations from 7 papers, 4 distinct outputs across 4 buckets, one shared input per bucket utils.py ran Apache-2.0 (permissive) ·
954d7736e2d968f3 · report
fixture: signature list_of_lists:list/1/list · returned list · level recorded 2026-09-24 712ccc81e62f75b9 · report
fixture: signature probs:ndarray/2/float, labels:list/1/str · returned list · level recorded 2026-09-24 df853d8826e7a476 · report
fixture: signature path:str/0/str, remove_header:bool/0/bool · returned list · level recorded 2026-09-24 cd2c2c1263e0882d · report
fixture: signature text:str/0/str · returned str · level recorded 2026-09-24 a9831bf0d050a65c · report
fixture: signature text:str/0/str · returned str · level recorded 2026-09-24 53e20baefb5f4cd1 · report
census: 11 implementations from 18 papers, 7 distinct outputs across 3 buckets, one shared input per bucket qs_kpa/baselines/tf_models.py ran Apache-2.0 (permissive) ·
a5f2db749ea77f58 · report
fixture: signature y:EagerTensor/1/float, preds:EagerTensor/1/float, margin:float/0/float · returned EagerTensor · level recorded 2026-09-24 6708df0bbbc13e34 · report
fixture: signature normed_feature:Tensor/2/float, label:Tensor/1/int · returned tuple · level recorded 2026-09-24 census: 5 implementations from 5 papers, 4 distinct outputs across 3 buckets, one shared input per bucket qs_kpa/losses/multiple_negatives_ranking_loss.py ran · honoured contract fingerprinted Apache-2.0 (permissive) ·
8ca90d5388372be5 · report
fixture: signature a:Tensor/2/float, b:Tensor/2/float · returned Tensor · level recorded 2026-09-24 afa18317f44e63ad · report
fixture: signature a:Tensor/2/float, b:Tensor/2/float · returned Tensor · level recorded 2026-09-24 e65cfdc18cd8c48f · report
fixture: signature emissions:Tensor/3/float, tags:Tensor/2/int, mask:Tensor/2/int, reduction:str/0/str · returned Tensor · level recorded 2026-09-24 7d40208cd5d0da23 · report
68faf3c1193d95e7 · report
fixture: returned Encoder · level recorded 2026-09-22 9d6930c9ef7f273b · report
fixture: returned Model · level recorded 2026-09-22 717e1f9d4eae4141 · report
fixture: returned Multi_Head_Attention · level recorded 2026-09-22 c15e979783157fb0 · report
fixture: returned Position_wise_Feed_Forward · level recorded 2026-09-22 ce8217cf9489c436 · report
fixture: returned Positional_Encoding · level recorded 2026-09-22 efea4e3ee53c36c9 · report
fixture: signature Q:Tensor/3/float, K:Tensor/3/float, V:Tensor/3/float, scale:float/0/float · returned Tensor · level recorded 2026-09-22 6788010bd752077c · report
8afaa985063f723e · report
fixture: signature module_or_pth_file:callable/0/?, model:str/0/str · returned dict · level recorded 2026-09-24 e8ae42ea576c0232 · report
fixture: signature attention_mask:Tensor/2/int · returned Tensor · level recorded 2026-09-24 census: 9 implementations from 11 papers, 3 distinct outputs on one shared input cybertron/models/bart.py unverified Apache-2.0 (permissive) ·
8f0494a0f9a303b8 · report
Method as code Syntology
What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it. Siblings are assemblies lifted from this paper in the same batch for the same method: alternative readings kept side by side; none replaces or supersedes another.
12 assemblies on this paper: 12 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).
12 sibling assemblies for RoBERTa (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.
harvested Assembly for RoBERTa (graph method: RoBERTa). 4 bricks; individually on a fixture: 2 ran (of which 2 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified; every one of the 2 samples that ran constructed an object rather than computing a result. The assembly has not been run end-to-end.
one of 12 siblings
- Easyrec entry · level 0 · unverified ·
b2bd29f8a81790e2 - MLPLayer model · level 0 · unverified ·
7d7789da5378c84b - Pooler model · level 1 · ran ·
dba0b61c9d62f82f - Similarity helper · level 1 · ran ·
4d41003f996241f8
harvested Assembly for RoBERTa (graph method: RoBERTa). 2 bricks; individually on a fixture: 1 ran (of which 0 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 assembly has not been run end-to-end.
one of 12 siblings
- RoBertaCRF entry · level 0 · unverified ·
7d40208cd5d0da23 - CRF model · level 1 · ran ·
e65cfdc18cd8c48f
harvested Assembly for RoBERTa (graph method: RoBERTa). 2 bricks; individually on a fixture: 2 ran (of which 1 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- CustomRobertaForSequenceClassification entry · level 1 · ran ·
0196b04010ddfa00 - ClassificationHead model · level 1 · ran ·
7b8d60647337c2b1
harvested Assembly for RoBERTa (graph method: RoBERTa). 10 bricks; individually on a fixture: 8 ran (of which 4 constructed an object rather than computing a result; 8 with no instrument failure: 0 honoured, 0 violated, 8 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- BertModel entry · level 1 · ran ·
f43b4161b7bdba24 - BertEmbeddings model · level 1 · ran ·
c3af34b4f438eaea - BertEncoder model · level 0 · unverified ·
00ea7705507db813 - BertPooler model · level 1 · ran ·
4e755eff48bf0752 - BertLayer model · level 0 · unverified ·
f2ab856479b90e8a - BertAttention model · level 1 · ran ·
ad7f6e57dedf08aa - BertIntermediate model · level 1 · ran ·
6e60ffe93c5ec5fa - BertOutput model · level 1 · ran ·
5e5629cb8e93e004 - BertSelfAttention model · level 1 · ran ·
df8363b5430892a0 - BertSelfOutput model · level 1 · ran ·
fcf3cd282eb03efe
harvested Assembly for RoBERTa (graph method: RoBERTa). 6 bricks; individually on a fixture: 6 ran (of which 5 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- Model entry · level 1 · ran ·
9d6930c9ef7f273b - Encoder helper · level 1 · ran ·
68faf3c1193d95e7 - Positional_Encoding model · level 1 · ran ·
ce8217cf9489c436 - Multi_Head_Attention model · level 1 · ran ·
717e1f9d4eae4141 - Position_wise_Feed_Forward model · level 1 · ran ·
c15e979783157fb0 - Scaled_Dot_Product_Attention model · level 1 · ran ·
efea4e3ee53c36c9
harvested Assembly for RoBERTa (graph method: RoBERTa). 1 brick; individually on a fixture: 0 ran · 1 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- QuestionAnswering entry · level 0 · unverified ·
da835c893505251b
harvested Assembly for RoBERTa (graph method: RoBERTa). 4 bricks; individually on a fixture: 4 ran (of which 3 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- Encoder entry · level 1 · ran ·
68faf3c1193d95e7 - Multi_Head_Attention model · level 1 · ran ·
717e1f9d4eae4141 - Position_wise_Feed_Forward model · level 1 · ran ·
c15e979783157fb0 - Scaled_Dot_Product_Attention helper · level 1 · ran ·
efea4e3ee53c36c9
harvested Assembly for RoBERTa (graph method: RoBERTa). 1 brick; individually on a fixture: 0 ran · 1 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- RobertaForMultiLabelSequenceClassification entry · level 0 · unverified ·
b3e0c37e1e2b9503
harvested Assembly for RoBERTa (graph method: RoBERTa). 1 brick; individually on a fixture: 0 ran · 1 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- RobertaForMultiLabelSequenceClassification entry · level 0 · unverified ·
6788010bd752077c
harvested Assembly for RoBERTa (graph method: RoBERTa). 1 brick; individually on a fixture: 0 ran · 1 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- train entry · level 0 · unverified ·
92e80bcaa1f41216
harvested Assembly for RoBERTa (graph method: RoBERTa). 1 brick; individually on a fixture: 1 ran (of which 0 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) · 0 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- RobertaNER entry · level 1 · ran ·
152609ff366dfb9e
harvested Assembly for RoBERTa (graph method: RoBERTa). 1 brick; individually on a fixture: 1 ran (of which 0 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) · 0 unverified. The assembly has not been run end-to-end.
one of 12 siblings
- RobertaClassificationHead entry · level 1 · ran ·
17243795c9ab1ec3
Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="1907.11692") returns these assemblies with each brick's role and wiring (how to connect).
Tasks
Results from the paper archive 2025-07-28
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
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