EasyDistill: A Comprehensive Toolkit for Effective Knowledge Distillation of Large Language Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Chengyu, Yan, Junbing, Cai, Wenrui, Yue, Yuanhao, Huang, Jun
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912452652826624
author Wang, Chengyu
Yan, Junbing
Cai, Wenrui
Yue, Yuanhao
Huang, Jun
author_facet Wang, Chengyu
Yan, Junbing
Cai, Wenrui
Yue, Yuanhao
Huang, Jun
contents In this paper, we present EasyDistill, a comprehensive toolkit designed for effective black-box and white-box knowledge distillation (KD) of large language models (LLMs). Our framework offers versatile functionalities, including data synthesis, supervised fine-tuning, ranking optimization, and reinforcement learning techniques specifically tailored for KD scenarios. The toolkit accommodates KD functionalities for both System 1 (fast, intuitive) and System 2 (slow, analytical) models. With its modular design and user-friendly interface, EasyDistill empowers researchers and industry practitioners to seamlessly experiment with and implement state-of-the-art KD strategies for LLMs. In addition, EasyDistill provides a series of robust distilled models and KD-based industrial solutions developed by us, along with the corresponding open-sourced datasets, catering to a variety of use cases. Furthermore, we describe the seamless integration of EasyDistill into Alibaba Cloud's Platform for AI (PAI). Overall, the EasyDistill toolkit makes advanced KD techniques for LLMs more accessible and impactful within the NLP community.
format Preprint
id arxiv_https___arxiv_org_abs_2505_20888
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle EasyDistill: A Comprehensive Toolkit for Effective Knowledge Distillation of Large Language Models
Wang, Chengyu
Yan, Junbing
Cai, Wenrui
Yue, Yuanhao
Huang, Jun
Computation and Language
Artificial Intelligence
In this paper, we present EasyDistill, a comprehensive toolkit designed for effective black-box and white-box knowledge distillation (KD) of large language models (LLMs). Our framework offers versatile functionalities, including data synthesis, supervised fine-tuning, ranking optimization, and reinforcement learning techniques specifically tailored for KD scenarios. The toolkit accommodates KD functionalities for both System 1 (fast, intuitive) and System 2 (slow, analytical) models. With its modular design and user-friendly interface, EasyDistill empowers researchers and industry practitioners to seamlessly experiment with and implement state-of-the-art KD strategies for LLMs. In addition, EasyDistill provides a series of robust distilled models and KD-based industrial solutions developed by us, along with the corresponding open-sourced datasets, catering to a variety of use cases. Furthermore, we describe the seamless integration of EasyDistill into Alibaba Cloud's Platform for AI (PAI). Overall, the EasyDistill toolkit makes advanced KD techniques for LLMs more accessible and impactful within the NLP community.
title EasyDistill: A Comprehensive Toolkit for Effective Knowledge Distillation of Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.20888