MétaCan
Menu
Back to cohort

Revisiting Intermediate Layer Distillation for Compressing Language Models: An Overfitting Perspective

2023· article· en· W4386566583 on OpenAlexaff
Jongwoo Ko, Seungjoon Park, Minchan Jeong, Suk-Jin Hong, Euijai Ahn, Du-Seong Chang, Se-Young Yun

Bibliographic record

Venuenot available
Typearticle
Languageen
FieldComputer Science
TopicTopic Modeling
Canadian institutionsKootenay Association for Science & Technology
FundersInstitute for Information and Communications Technology PromotionMinistry of Science and ICT, South KoreaKorea Advanced Institute of Science and Technology
KeywordsOverfittingComputer scienceDistillationCode (set theory)Benchmark (surveying)Machine learningArtificial intelligenceConsistency (knowledge bases)Simple (philosophy)TransformerLayer (electronics)Language modelNatural language processingArtificial neural networkProgramming languageEngineeringChemistry

Abstract

fetched live from OpenAlex

Knowledge distillation (KD) is a highly promising method for mitigating the computational problems of pre-trained language models (PLMs).Among various KD approaches, Intermediate Layer Distillation (ILD) has been a de facto standard KD method with its performance efficacy in the NLP field.In this paper, we find that existing ILD methods are prone to overfitting to training datasets, although these methods transfer more information than the original KD.Next, we present the simple observations to mitigate the overfitting of ILD: distilling only the last Transformer layer and conducting ILD on supplementary tasks.Based on our two findings, we propose a simple yet effective consistency-regularized ILD (CR-ILD), which prevents the student model from overfitting the training dataset.Substantial experiments on distilling BERT on the GLUE benchmark and several synthetic datasets demonstrate that our proposed ILD method outperforms other KD techniques.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.003
metaresearch head score (Gemma)0.014
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: none
Teacher disagreement score0.006
Threshold uncertainty score0.015

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0030.014
Meta-epidemiology (narrow)0.0020.001
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0010.002
Science and technology studies0.0010.001
Scholarly communication0.0020.005
Open science0.0030.004
Research integrity0.0010.005
Insufficient payload (model declined to judge)0.0040.003

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.066
GPT teacher head0.336
Teacher spread0.270 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations4
Published2023
Admission routes1
Has abstractyes

Explore more

Same topicTopic ModelingFrench-language works237,207