# Understanding the Effect of Data Augmentation on Knowledge Distillation

arXiv: https://arxiv.org/abs/2305.12565 · HTML: https://mlsanity.com/paper/2305.12565/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in General knowledge distillation methods, one of 12 directions in Knowledge distillation.
- Builds on 21 of its 36 references that are in the map.
- First posted on arXiv 21 May 2023.

## Placed in

- Knowledge distillation / [General knowledge distillation methods](https://mlsanity.com/distill/general-knowledge-distillation-methods/)

## Builds on (most cited first, 21 of 21)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference](https://mlsanity.com/paper/1704.05426/) (2017)
- [BioBERT: a pre-trained biomedical language representation model for biomedical text mining](https://mlsanity.com/paper/1901.08746/) (2019)
- [No Language Left Behind: Scaling Human-Centered Machine Translation](https://mlsanity.com/paper/2207.04672/) (2022)
- [TinyBERT: Distilling BERT for Natural Language Understanding](https://mlsanity.com/paper/1909.10351/) (2019)
- [EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks](https://mlsanity.com/paper/1901.11196/) (2019)
- [Patient Knowledge Distillation for BERT Model Compression](https://mlsanity.com/paper/1908.09355/) (2019)
- [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://mlsanity.com/paper/2004.02984/) (2020)
- [GPT3Mix: Leveraging Large-scale Language Models for Text Augmentation](https://mlsanity.com/paper/2104.08826/) (2021)
- [Distilling Task-Specific Knowledge from BERT into Simple Neural Networks](https://mlsanity.com/paper/1903.12136/) (2019)
- [MixKD: Towards Efficient Distillation of Large-scale Language Models](https://mlsanity.com/paper/2011.00593/) (2020)
- [What Makes a "Good" Data Augmentation in Knowledge Distillation -- A Statistical Perspective](https://mlsanity.com/paper/2012.02909/) (2020)
- [FlipDA: Effective and Robust Data Augmentation for Few-Shot Learning](https://mlsanity.com/paper/2108.06332/) (2021)
- [TreeMix: Compositional Constituency-based Data Augmentation for Natural Language Understanding](https://mlsanity.com/paper/2205.06153/) (2022)
- [Learning to Augment for Data-Scarce Domain BERT Knowledge Distillation](https://mlsanity.com/paper/2101.08106/) (2021)
- [Augmentation with Projection: Towards an Effective and Efficient Data Augmentation Paradigm for Distillation](https://mlsanity.com/paper/2210.11768/) (2022)

## Built on by (most cited first, 2 of 2)

- [Defining a New NLP Playground](https://mlsanity.com/paper/2310.20633/) (2023)
- [Diffusion-Assisted Distillation for Self-Supervised Graph Representation Learning with MLPs](https://mlsanity.com/paper/2510.04241/) (2025)

## Abstract (arXiv)

Knowledge distillation (KD) requires sufficient data to transfer knowledge from large-scale teacher models to small-scale student models. Therefore, data augmentation has been widely used to mitigate the shortage of data under specific scenarios. Classic data augmentation techniques, such as synonym replacement and k-nearest-neighbors, are initially designed for fine-tuning. To avoid severe semantic shifts and preserve task-specific labels, those methods prefer to change only a small proportion of tokens (e.g., changing 10% tokens is generally the best option for fine-tuning). However, such data augmentation methods are sub-optimal for knowledge distillation since the teacher model could provide label distributions and is more tolerant to semantic shifts. We first observe that KD prefers as much data as possible, which is different from fine-tuning that too much data will not gain more performance. Since changing more tokens leads to more semantic shifts, we use the proportion of changed tokens to reflect semantic shift degrees. Then we find that KD prefers augmented data with a larger semantic shift degree (e.g., changing 30% tokens is generally the best option for KD) than fine-tuning (changing 10% tokens). Besides, our findings show that smaller datasets prefer larger degrees until the out-of-distribution problem occurs (e.g., datasets with less than 10k inputs may prefer the 50% degree, and datasets with more than 100k inputs may prefer the 10% degree). Our work sheds light on the preference difference in data augmentation between fine-tuning and knowledge distillation and encourages the community to explore KD-specific data augmentation methods.