Challenge: Knowledge distillation (KD) is the preliminary step for training non-autoregressive translation models, but it can lose important information for translating low-frequency words.
Approach: They propose a knowledge distillation method which trains NAT student on external monolingual data with AT teacher trained on the original bilingual data.
Outcome: Extensive experiments on eight WMT benchmarks show that monolingual KD outperforms the standard KD by improving low-frequency word translation without introducing any computational cost.

Similar Papers

Rejuvenating Low-Frequency Words: Making the Most of Parallel Data in Non-Autoregressive Translation (2021.acl-long)

Copied to clipboard

Challenge: Knowledge distillation (KD) is commonly used to construct synthetic data for training non-autoregressive translation models.
Approach: They propose to use knowledge distillation to generate training data for non-autoregressive translation models by leveraging pretraining.
Outcome: The proposed approach achieves 28.2 and 33.9 BLEU points on the WMT14 English-German and WMT16 Romanian-English datasets.
Improving Non-autoregressive Neural Machine Translation with Monolingual Data (2020.acl-main)

Copied to clipboard

Challenge: Neural machine translation is usually done via knowledge distillation from an autoregressive (AR) model.
Approach: They leverage large monolingual corpora to improve the NAR model's performance by transferring the autoregressive model' s generalization ability while preventing overfitting.
Outcome: The proposed methods on the WMT14 En-De and WMT16 En-Ro news translation tasks show that monolingual data augmentation improves the NAR model to approach the teacher AR model’s performance.
Revisiting Knowledge Distillation for Autoregressive Language Models (2024.acl-long)

Copied to clipboard

Challenge: Autoregressive language models (LMs) are expensive and memory intensive, preventing the development of industrial applications.
Approach: They propose an adaptive teaching approach to improve the KD of autoregressive language models by distilling knowledge into a small student model.
Outcome: The proposed method can achieve consistent and significant performance gains across all model types and sizes.
Towards Zero-Shot Knowledge Distillation for Natural Language Processing (2021.emnlp-main)

Copied to clipboard

Challenge: Knowledge distillation (KD) is a common knowledge transfer algorithm used for model compression across a variety of deep learning based natural language processing (NLP) solutions.
Approach: They propose to use teacher training data for model compression . they investigate six tasks and find they can achieve between 75% and 92% of the teacher’s classification score while compressing the model 30 times.
Outcome: The proposed solution achieves between 75% and 92% of the teacher’s classification score while compressing the model 30 times.
Dual-Space Knowledge Distillation for Large Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Existing large language models (LLMs) have strong generalization abilities due to their huge model capacities.
Approach: They propose a dual-space knowledge distillation framework that unifies the output spaces of the two models for KD.
Outcome: The proposed framework outperforms existing white-box KD frameworks on task-agnostic instruction-following benchmarks and can automatically align representations of two models with different vocabularies.
Towards Understanding and Improving Knowledge Distillation for Neural Machine Translation (2023.acl-long)

Copied to clipboard

Challenge: Existing knowledge distillation techniques for neural machine translation lack special treatment on the top-1 information, which is limiting the potential of KD.
Approach: They propose a method to distill knowledge from top-1 predictions of teachers and a technique to infuse more additional knowledge by distilling on the data without ground-truth targets.
Outcome: The proposed method outperforms the vanilla word-level KD and outperfies the existing methods on three different students with different capacity gaps.
Dual-teacher Knowledge Distillation for Low-frequency Word Translation (2024.findings-emnlp)

Copied to clipboard

Challenge: Neural machine translation models are trained on parallel corpora with unbalanced word frequency distribution, resulting in high-frequency words being ignored.
Approach: They propose to employ a low-frequency teacher model that excels in translating low- frequency words to guide the learning of the student model.
Outcome: The proposed method achieves +0.64 BLEU improvements over the state-of-the-art method on the low-frequency translation task while maintaining the translation quality of high-frequency words.
Meta-KD: A Meta Knowledge Distillation Framework for Language Model Compression across Domains (2021.acl-long)

Copied to clipboard

Challenge: Pre-trained language models have been successful in NLP tasks, but their large size and long inference time limit their deployment in real-time applications.
Approach: They propose a meta-teacher model that captures transferable knowledge across domains and passes it to students.
Outcome: The proposed model can distill large teacher models into small student models with guidance from the meta-teacher.
Generation-Distillation for Efficient Natural Language Understanding in Low-Data Settings (D19-61)

Copied to clipboard

Challenge: Recent research points to knowledge distillation as a potential solution for NLU tasks.
Approach: They propose a training approach that distills large finetuned LMs into a small network using unlabeled training examples.
Outcome: The proposed approach outperforms BERT training approaches while using 300 times fewer parameters.
Beyond the Mode: Sequence-Level Distillation of Multilingual Translation Models for Low-Resource Language Pairs (2025.findings-naacl)

Copied to clipboard

Challenge: Existing multilingual pre-trained models for low-resource languages have outperformed those trained from scratch for low resources due to high hardware requirements.
Approach: They propose to use beam search to decode the whole output distribution of the teacher to improve student learning.
Outcome: The proposed methods improve student model performance and reduce gender bias amplification common to beam search based methods.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations