Saved in:
Bibliographic Details
Main Author: Setiawan, Hendra
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2305.12057
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911914670424064
author Setiawan, Hendra
author_facet Setiawan, Hendra
contents We propose utilizing n-best reranking to enhance Sequence-Level Knowledge Distillation (Kim and Rush, 2016) where we extract pseudo-labels for student model's training data from top n-best hypotheses and leverage a diverse set of models with different inductive biases, objective functions or architectures, including some publicly-available large language models, to pick the highest-quality hypotheses as labels. The effectiveness of our proposal is validated through experiments on the WMT'21 German-English and Chinese-English translation tasks. Our results demonstrate that utilizing pseudo-labels generated by our n-best reranker leads to a significantly more accurate student model. In fact, our best student model achieves comparable accuracy to a large translation model from (Tran et al., 2021) with 4.7 billion parameters, while having two orders of magnitude fewer parameters.
format Preprint
id arxiv_https___arxiv_org_abs_2305_12057
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Accurate Knowledge Distillation with n-best Reranking
Setiawan, Hendra
Computation and Language
We propose utilizing n-best reranking to enhance Sequence-Level Knowledge Distillation (Kim and Rush, 2016) where we extract pseudo-labels for student model's training data from top n-best hypotheses and leverage a diverse set of models with different inductive biases, objective functions or architectures, including some publicly-available large language models, to pick the highest-quality hypotheses as labels. The effectiveness of our proposal is validated through experiments on the WMT'21 German-English and Chinese-English translation tasks. Our results demonstrate that utilizing pseudo-labels generated by our n-best reranker leads to a significantly more accurate student model. In fact, our best student model achieves comparable accuracy to a large translation model from (Tran et al., 2021) with 4.7 billion parameters, while having two orders of magnitude fewer parameters.
title Accurate Knowledge Distillation with n-best Reranking
topic Computation and Language
url https://arxiv.org/abs/2305.12057