Papers › DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with...

DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing

18 Nov 2021arXiv:2111.09543archive 2025-07-28

Pengcheng He, Jianfeng Gao, Weizhu Chen

This paper presents a new pre-trained language model, DeBERTaV3, which improves the original DeBERTa model by replacing mask language modeling (MLM) with replaced token detection (RTD), a more sample-efficient pre-training task. Our analysis shows that vanilla embedding sharing in ELECTRA hurts training efficiency and model performance. This is because the training losses of the discriminator and the generator pull token embeddings in different directions, creating the "tug-of-war" dynamics. We thus propose a new gradient-disentangled embedding sharing method that avoids the tug-of-war dynamics, improving both training efficiency and the quality of the pre-trained model. We have pre-trained DeBERTaV3 using the same settings as DeBERTa to demonstrate its exceptional performance on a wide range of downstream natural language understanding (NLU) tasks. Taking the GLUE benchmark with eight tasks as an example, the DeBERTaV3 Large model achieves a 91.37% average score, which is 1.37% over DeBERTa and 1.91% over ELECTRA, setting a new state-of-the-art (SOTA) among the models with a similar structure. Furthermore, we have pre-trained a multi-lingual model mDeBERTa and observed a larger improvement over strong baselines compared to English models. For example, the mDeBERTa Base achieves a 79.8% zero-shot cross-lingual accuracy on XNLI and a 3.6% improvement over XLM-R Base, creating a new SOTA on this benchmark. We have made our pre-trained models and inference code publicly available at https://github.com/microsoft/DeBERTa.

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

Syntology 7 code samples harvested from 1 repository linked to this paper: 6 ran (of which 0 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) · 1 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code; here there are none.

By repository: official repository: 7 samples from 1 repository, 6 ran (6 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.

Levels written on 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

microsoft/DeBERTa officialmentioned in papermentioned on GitHubpytorchMITSyntology: 7 samples, 6 ran (0 instrument-failed) · 1 unverified report
dashenzi721/hra mentioned on GitHubpytorchSyntology: not harvested report
stareru/csqa_debertav3 mentioned on GitHubpytorchSyntology: not harvested 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.

Code Syntology ran Syntology

7 samples harvested: 6 ran (of which 0 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) · 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.

6ran
1unverified

Licence: 0 of the 7 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 microsoft/DeBERTa. “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.

microsoft/DeBERTa official repository 7 samples, 6 ran (0 instrument-failed) · 1 unverified
build_relative_position in 17 papers DeBERTa/deberta/da_utils.py ran MIT (permissive) · ac8689e2d4eb4c75 · report fixture: signature query_size:int/0/int, key_size:int/0/int, bucket_size:int/0/int, max_position:int/0/int, device:NoneType/0/? · returned Tensor · level recorded 2026-09-25
join_chunks DeBERTa/apps/_utils.py ran MIT (permissive) · f7b253455c817d10 · report fixture: signature chunks:list/1/Tensor, max_len:NoneType/0/? · returned Tensor · level recorded 2026-09-25
load_vocab in 107 papers DeBERTa/deberta/cache_utils.py ran MIT (permissive) · 82c76e446399db0c · report fixture: signature vocab_path:str/0/str, vocab_type:str/0/str, pretrained_id:NoneType/0/?, tag:NoneType/0/?, no_cache:bool/0/bool, cache_dir:str/0/str · returned tuple · level recorded 2026-09-25
make_log_bucket_dict DeBERTa/deberta/da_utils.py ran MIT (permissive) · 792471bf746ab83e · report fixture: signature bucket_size:int/0/int, max_position:int/0/int, device:NoneType/0/? · returned Tensor · level recorded 2026-09-25
make_log_bucket_position in 15 papers DeBERTa/deberta/da_utils.py ran fingerprinted MIT (permissive) · 7af87f19f9630217 · report fixture: signature relative_pos:Tensor/2/int, bucket_size:int/0/int, max_position:int/0/int · returned Tensor · level recorded 2026-09-25
merge_distributed DeBERTa/apps/_utils.py ran MIT (permissive) · c734e14b42bb19f0 · report fixture: signature data_list:list/1/Tensor, max_len:NoneType/0/? · returned Tensor · level recorded 2026-09-25
download_asset DeBERTa/deberta/cache_utils.py unverified MIT (permissive) · ce127028071b7db8 · report

Tasks

Language ModelingLanguage ModellingNatural Language InferenceNatural Language UnderstandingQuestion AnsweringXLM-R

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Natural Language Inference MRPC DeBERTaV3large Acc 92.2 #1 of 1 Archive leaderboard report
Natural Language Inference QNLI DeBERTaV3large Accuracy 96% #8 of 43 Archive leaderboard report
Natural Language Inference RTE DeBERTaV3large Accuracy 92.7% #7 of 90 Archive leaderboard report
Question Answering SWAG DeBERTaV3large Accuracy 93.4 #1 of 1 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

AdamAttentionAttention DropoutDeBERTaDense ConnectionsDropoutELECTRALayer NormalizationLinear LayerLinear Warmup With Linear DecayMulti-Head AttentionResidual ConnectionSoftmaxWeight DecayWordPieceXLM-R

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