REVIEW 30 cited by
TinyBERT: Distilling BERT for Natural Language Understanding
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
TinyBERT: Distilling BERT for Natural Language Understanding
read the original abstract
Language model pre-training, such as BERT, has significantly improved the performances of many natural language processing tasks. However, pre-trained language models are usually computationally expensive, so it is difficult to efficiently execute them on resource-restricted devices. To accelerate inference and reduce model size while maintaining accuracy, we first propose a novel Transformer distillation method that is specially designed for knowledge distillation (KD) of the Transformer-based models. By leveraging this new KD method, the plenty of knowledge encoded in a large teacher BERT can be effectively transferred to a small student Tiny-BERT. Then, we introduce a new two-stage learning framework for TinyBERT, which performs Transformer distillation at both the pretraining and task-specific learning stages. This framework ensures that TinyBERT can capture he general-domain as well as the task-specific knowledge in BERT. TinyBERT with 4 layers is empirically effective and achieves more than 96.8% the performance of its teacher BERTBASE on GLUE benchmark, while being 7.5x smaller and 9.4x faster on inference. TinyBERT with 4 layers is also significantly better than 4-layer state-of-the-art baselines on BERT distillation, with only about 28% parameters and about 31% inference time of them. Moreover, TinyBERT with 6 layers performs on-par with its teacher BERTBASE.
Forward citations
Cited by 30 Pith papers
-
Language Models are Few-Shot Learners
GPT-3 shows that scaling an autoregressive language model to 175 billion parameters enables strong few-shot performance across diverse NLP tasks via in-context prompting without fine-tuning.
-
ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
ELECTRA replaces masked language modeling with replaced token detection, yielding contextual representations that outperform BERT at equal compute and match larger models like RoBERTa with far less compute.
-
Toward Calibrated, Fair, and accurate Deepfake Detection
Face-Feature Tuning is a label-free logit remapping method that reduces FPR/TPR gaps across groups in deepfake detection while preserving overall accuracy.
-
Bounded Behavioral Indistinguishability for Black-Box LLM Distillation
Introduces (ε,q,t,A)-behavioral indistinguishability and shows via Qwen/Llama experiments that LoRA distillation boosts semantic similarity but leaves detectable behavioral differences under adversarial evaluation.
-
MTA: Multi-Granular Trajectory Alignment for Large Language Model Distillation
MTA improves LLM knowledge distillation by aligning representations along layer-wise trajectories with adaptive granularity from words to phrases using dynamic structural and hidden representation alignment losses.
-
SRA: Span Representation Alignment for Large Language Model Distillation
SRA reframes CTKD by aligning attention-weighted span centers of mass in a multi-particle system model with geometric regularization and span logit distillation, claiming consistent outperformance over baselines.
-
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
T5 casts all NLP tasks as text-to-text generation, systematically explores pre-training choices, and reaches strong performance on summarization, QA, classification and other tasks via large-scale training on the Colo...
-
Data Provenance for Image Auto-Regressive Generation
A post-hoc detection framework exploits generation-induced patterns in autoregressive image outputs to enable provenance tracing across multiple IAR models without altering the generation process.
-
Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models
Empirical benchmarks show distribution similarity between adaptation and pretraining data increases practical privacy leakage in DP-adapted LLMs at fixed theoretical guarantees, with LoRA providing strongest protectio...
-
Do Transformers Need Three Projections? Systematic Study of QKV Variants
Q-K=V projection sharing in transformers matches standard QKV performance with 50% KV cache reduction and combines with GQA/MQA for up to 96.9% reduction across vision and language tasks.
-
Combining pre-trained models via localized model averaging
Localized model averaging with covariate-dependent weights achieves asymptotic optimality and weight consistency for combining pre-trained models under a general loss framework.
-
Modular Representation Compression: Adapting LLMs for Efficient and Effective Recommendations
LLMs exhibit mid-layer representation advantage for recommendations; MARC compresses representations modularly to reduce costs while improving performance, as shown in a large-scale online advertising deployment.
-
BiSpikCLM: A Spiking Language Model integrating Softmax-Free Spiking Attention and Spike-Aware Alignment Distillation
BiSpikCLM is the first fully binary spiking MatMul-free causal language model that matches ANN performance on generation tasks using only 4-6 percent of the compute via softmax-free spiking attention and spike-aware d...
-
A Comparative Study of Semantic Log Representations for Software Log-based Anomaly Detection
QTyBERT matches or exceeds BERT-based log anomaly detection effectiveness while reducing embedding generation time to near static word embedding levels.
-
A Metamorphic Testing Perspective on Knowledge Distillation for Language Models of Code: Does the Student Deeply Mimic the Teacher?
Student models distilled from code language models often fail to deeply mimic teachers, showing up to 62% behavioral discrepancies and 285% worse drops under attacks that accuracy metrics miss.
-
DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing
DeBERTaV3 improves DeBERTa by switching to replaced token detection pre-training and using gradient-disentangled embedding sharing, reaching 91.37% on GLUE and new SOTA on XNLI zero-shot.
-
Knowledge Distillation in Iterative Generative Models for Improved Sampling Speed
Denoising Student distills the multi-step denoising process of score-based and diffusion models into a single forward pass, matching GAN sampling speed while producing comparable sample quality on CIFAR-10, CelebA, an...
-
Vitality-Aware Compression for Efficient Image-to-Shape Diffusion Transformers
Introduces vitality-aware compression for image-to-3D DiT models via structured pruning, adaptive quantization, and fine-tuning, claiming 66% size reduction with comparable fidelity.
-
Unified Multi-Task Relevance Modeling for E-Commerce: Comparing Task Routing Architectures Across LLMs and Cross-Encoders
A multi-head private-layer ensemble in a unified multi-task setup reaches 89.96% accuracy on 453K e-commerce examples and improves low-resource tasks by up to 14% while revealing encoder-decoder asymmetry in task iden...
-
Minimizing the Hidden Cost of Scales: Graph-Guided Ultra-Low-Bit Quantization for Large Language Models
SAGE-PTQ is a graph-guided ultra-low-bit PTQ framework that achieves 1.03 average weight bits and 0.004 scaling bits per matrix on LLMs while reporting lower perplexity and memory use than BiLLM and PB-LLM.
-
When Mean CE Fails: Median CE Can Better Track Language Model Quality
Median cross-entropy tracks language model task performance more reliably than mean cross-entropy during synthetic fact-learning SFT and top-K distillation.
-
Fast Tensorization of Neural Networks via Slice-wise Feature Distillation
A slice-wise feature distillation framework for independent tensorization of neural network slices to achieve scalable compression with reduced fine-tuning costs.
-
MTA: Multi-Granular Trajectory Alignment for Large Language Model Distillation
MTA is a distillation method that aligns teacher-student LLM representations along their transformation trajectories using layer-adaptive granularities and dynamic structural plus hidden representation alignment losses.
-
SRA: Span Representation Alignment for Large Language Model Distillation
SRA reframes cross-tokenizer LLM distillation as alignment of attention-weighted span centers of mass in a multi-particle dynamical system and reports consistent gains over prior CTKD baselines.
-
TabEmb: Joint Semantic-Structure Embedding for Table Annotation
TabEmb decouples LLM-based semantic column embeddings from graph-based structural modeling to produce joint representations that improve table annotation tasks.
-
ESsEN: Training Compact Discriminative Vision-Language Transformers in a Low-Resource Setting
ESsEN is a parameter-efficient two-tower vision-language transformer that matches larger models on discriminative tasks after training end-to-end with limited data and resources.
-
A Transformer-Based Cross-Platform Analysis of Public Discourse on the 15-Minute City Paradigm
Benchmarks five compressed transformer models for multi-platform sentiment classification on 15-minute city discourse, reporting DistilRoBERTa highest F1 of 0.8292 and platform-specific performance differences.
-
Model Compression vs. Adversarial Robustness: An Empirical Study on Language Models for Code
Empirical tests show compressed code language models retain task performance but suffer markedly lower robustness under four standard adversarial attacks.
-
TALAS: Teacher-Anchored Layer Alignment with Adaptive Sharpness-Aware Minimization for Embedding Distillation
TALAS is a knowledge distillation method that selectively aligns upper student layers to teacher sentence embeddings, propagates knowledge top-down via relational constraints in lower layers, and uses ASAM to seek fla...
-
Will LLMs Scaling Hit the Wall? Breaking Barriers via Distributed Resources on Massive Edge Devices
Position paper claiming that distributed training across massive edge devices can overcome data depletion and centralized compute monopolies in LLM scaling.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.