Papers › How to Distill your BERT: An Empirical Study on the Impact of Weight Initialisation...

How to Distill your BERT: An Empirical Study on the Impact of Weight Initialisation and Distillation Objectives

24 May 2023arXiv:2305.15032archive 2025-07-28

Xinpeng Wang, Leonie Weissweiler, Hinrich Schütze, Barbara Plank

Recently, various intermediate layer distillation (ILD) objectives have been shown to improve compression of BERT models via Knowledge Distillation (KD). However, a comprehensive evaluation of the objectives in both task-specific and task-agnostic settings is lacking. To the best of our knowledge, this is the first work comprehensively evaluating distillation objectives in both settings. We show that attention transfer gives the best performance overall. We also study the impact of layer choice when initializing the student from the teacher layers, finding a significant impact on the performance in task-specific distillation. For vanilla KD and hidden states transfer, initialisation with lower layers of the teacher gives a considerable improvement over higher layers, especially on the task of QNLI (up to an absolute percentage change of 17.8 in accuracy). Attention transfer behaves consistently under different initialisation settings. We release our code as an efficient transformer-based model distillation framework for further studies.

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 11 code samples harvested from 1 repository linked to this paper: 6 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 1 where Syntology's instrument failed) · 5 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 11 samples from 1 repository, 6 ran (5 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.

Levels written between 2026-09-22 and 2026-09-26 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

mainlp/how-to-distill-your-bert officialmentioned in paperpytorchMITSyntology: 11 samples, 6 ran (1 instrument-failed) · 5 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.

Code Syntology ran Syntology

11 samples harvested: 6 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 1 where Syntology's instrument failed) · 5 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.

1ran · our draft was wrong
5ran
5unverified

Licence: 0 of the 11 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 mainlp/how-to-distill-your-bert. “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. 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.

mainlp/how-to-distill-your-bert official repository 11 samples, 6 ran (1 instrument-failed) · 5 unverified
att_val_kl task_agnostic_distillation/methods/feature_distill.py ran MIT (permissive) · 0888d8eb55277184 · report fixture: signature student_atts:list/1/Tensor, student_qkv:list/1/tuple, teacher_atts:list/1/Tensor, teacher_qkv:list/1/tuple, layer_selection:str/0/str · returned tuple · level recorded 2026-09-26
cosine_similarity in 120 papers
census: 26 implementations from 30 papers, 11 distinct outputs across 5 buckets, one shared input per bucket
task_agnostic_distillation/methods/pear_loss.py ran fingerprinted MIT (permissive) · c5070fa84fd2b490 · report fixture: signature a:Tensor/2/float, b:Tensor/2/float, eps:float/0/float · returned Tensor · level recorded 2026-09-26
get_valid_dataloader task_agnostic_distillation/run_pretraining.py ran MIT (permissive) · 76a475fa13f37cec · report fixture: signature args:Args/0/?, dataset:SimpleDataset/0/? · returned generator · level recorded 2026-09-22
inter_class_relation in 7 papers task_agnostic_distillation/methods/pear_loss.py ran fingerprinted MIT (permissive) · cda3d1271f46dae7 · report fixture: signature y_s:Tensor/2/float, y_t:Tensor/2/float · returned Tensor · level recorded 2026-09-26
pearson_correlation in 15 papers task_agnostic_distillation/methods/pear_loss.py ran fingerprinted MIT (permissive) · 78c3c61943df0cc0 · report fixture: signature a:Tensor/2/float, b:Tensor/2/float, eps:float/0/float · returned Tensor · level recorded 2026-09-26
task_specific_config task_specific_distillation/experiments.py ran · our draft was wrong MIT (permissive) · de41371cecbab67c · report fixture: signature config:dict/0/?, task:str/0/str, epoch:int/0/int · returned dict · level recorded 2026-09-22
att_kl task_agnostic_distillation/methods/feature_distill.py unverified MIT (permissive) · 36f7375b15089e34 · report
bias_gelu in 7 papers task_agnostic_distillation/pretraining/modeling.py unverified MIT (permissive) · f114723b758d8789 · report
data_aug task_agnostic_distillation/methods/feature_distill.py unverified MIT (permissive) · e4c865c0ef8d62b4 · report
f_gelu task_agnostic_distillation/pretraining/modeling.py unverified MIT (permissive) · 2fd4f78b58e81a10 · report
get_deepspeed_config task_agnostic_distillation/pretraining/modeling.py unverified MIT (permissive) · 87b054b1860fbd9e · report

Tasks

Knowledge Distillation

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

AdamAttentionAttention DropoutBERTDense ConnectionsDropoutKnowledge DistillationLayer NormalizationLinear LayerLinear Warmup With Linear DecayMulti-Head AttentionResidual ConnectionSoftmaxWeight DecayWordPiece

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