# Sampling and Filtering of Neural Machine Translation Distillation Data

arXiv: https://arxiv.org/abs/2104.00664 · HTML: https://mlsanity.com/paper/2104.00664/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 10 of its 22 references that are in the map.
- First posted on arXiv 1 Apr 2021.

## Builds on (most cited first, 10 of 10)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing](https://mlsanity.com/paper/1808.06226/) (2018)
- [A Call for Clarity in Reporting BLEU Scores](https://mlsanity.com/paper/1804.08771/) (2018)
- [Marian: Fast Neural Machine Translation in C++](https://mlsanity.com/paper/1804.00344/) (2018)
- [Multilingual Neural Machine Translation with Knowledge Distillation](https://mlsanity.com/paper/1902.10461/) (2019)
- [Imitation Attacks and Defenses for Black-box Machine Translation Systems](https://mlsanity.com/paper/2004.15015/) (2020)
- [Microsoft Translator at WMT 2019: Towards Large-Scale Document-Level Neural Machine Translation](https://mlsanity.com/paper/1907.06170/) (2019)
- [Ensemble Distillation for Neural Machine Translation](https://mlsanity.com/paper/1702.01802/) (2017)
- [Announcing CzEng 2.0 Parallel Corpus with over 2 Gigawords](https://mlsanity.com/paper/2007.03006/) (2020)
- [Distill, Adapt, Distill: Training Small, In-Domain Models for Neural Machine Translation](https://mlsanity.com/paper/2003.02877/) (2020)

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

- [KD4MT: A Survey of Knowledge Distillation for Machine Translation](https://mlsanity.com/paper/2602.15845/) (2026)
- [Stolen Subwords: Importance of Vocabularies for Machine Translation Model Stealing](https://mlsanity.com/paper/2401.16055/) (2024)

## Abstract (arXiv)

In most of neural machine translation distillation or stealing scenarios, the goal is to preserve the performance of the target model (teacher). The highest-scoring hypothesis of the teacher model is commonly used to train a new model (student). If reference translations are also available, then better hypotheses (with respect to the references) can be upsampled and poor hypotheses either removed or undersampled. This paper explores the importance sampling method landscape (pruning, hypothesis upsampling and undersampling, deduplication and their combination) with English to Czech and English to German MT models using standard MT evaluation metrics. We show that careful upsampling and combination with the original data leads to better performance when compared to training only on the original or synthesized data or their direct combination.