@inproceedings{chenglin-etal-2024-mixed,
title = "Mixed Distillation Helps Smaller Language Models Reason Better",
author = "Chenglin, Li and
Chen, Qianglong and
Li, Liangyue and
Wang, Caiyu and
Tao, Feng and
Li, Yicheng and
Chen, Zulong and
Zhang, Yin",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-emnlp.91/",
doi = "10.18653/v1/2024.findings-emnlp.91",
pages = "1673--1690",
abstract = "As large language models (LLMs) have demonstrated impressive multiple step-by-step reasoning capabilities in recent natural language processing (NLP) reasoning tasks, many studies are interested in distilling reasoning abilities into smaller language models (SLMs) via fine-tuning. Previous distillation methods usually utilize the capabilities of LLMs to generate chain-of-thought (CoT) samples to teach SLMs. However, this distillation approach performs poorly in certain scenarios due to the limitations of CoT. In this work, we introduce a novel \textbf{Mixed Distillation} (MD) framework, distilling multiple step-by-step reasoning abilities into SLMs. First, we leverage LLMs to generate multiple step-by-step reasoning rationales by sampling automatically. Then, we create high-quality, well-balanced mixed thought data and design a novel multi-task loss to help SLMs better learn and adaptively activate multiple step-by-step reasoning. Our extensive experiments demonstrate that MD enhances both single-path (using either CoT or PoT) and multi-path (using both CoT and PoT) reasoning abilities of SLMs during inference across reasoning tasks. Notably, a single model generated by MD exceeds the comprehensive performance of an ensemble of two individual CoT and PoT distilled models. Mistral-7B using MD can achieve remarkable improvements of 87.5{\%}, 74.0{\%} and 77.1{\%} on SVAMP, GSM8K and ASDIV, respectively, outperforming the teacher model, GPT-3.5-Turbo. We hope our work provides insight into SLMs' multiple step-by-step reasoning abilities."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chenglin-etal-2024-mixed">
<titleInfo>
<title>Mixed Distillation Helps Smaller Language Models Reason Better</title>
</titleInfo>
<name type="personal">
<namePart type="given">Li</namePart>
<namePart type="family">Chenglin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qianglong</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liangyue</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Caiyu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Feng</namePart>
<namePart type="family">Tao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yicheng</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zulong</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yaser</namePart>
<namePart type="family">Al-Onaizan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Nung</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>As large language models (LLMs) have demonstrated impressive multiple step-by-step reasoning capabilities in recent natural language processing (NLP) reasoning tasks, many studies are interested in distilling reasoning abilities into smaller language models (SLMs) via fine-tuning. Previous distillation methods usually utilize the capabilities of LLMs to generate chain-of-thought (CoT) samples to teach SLMs. However, this distillation approach performs poorly in certain scenarios due to the limitations of CoT. In this work, we introduce a novel Mixed Distillation (MD) framework, distilling multiple step-by-step reasoning abilities into SLMs. First, we leverage LLMs to generate multiple step-by-step reasoning rationales by sampling automatically. Then, we create high-quality, well-balanced mixed thought data and design a novel multi-task loss to help SLMs better learn and adaptively activate multiple step-by-step reasoning. Our extensive experiments demonstrate that MD enhances both single-path (using either CoT or PoT) and multi-path (using both CoT and PoT) reasoning abilities of SLMs during inference across reasoning tasks. Notably, a single model generated by MD exceeds the comprehensive performance of an ensemble of two individual CoT and PoT distilled models. Mistral-7B using MD can achieve remarkable improvements of 87.5%, 74.0% and 77.1% on SVAMP, GSM8K and ASDIV, respectively, outperforming the teacher model, GPT-3.5-Turbo. We hope our work provides insight into SLMs’ multiple step-by-step reasoning abilities.</abstract>
<identifier type="citekey">chenglin-etal-2024-mixed</identifier>
<identifier type="doi">10.18653/v1/2024.findings-emnlp.91</identifier>
<location>
<url>https://aclanthology.org/2024.findings-emnlp.91/</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>1673</start>
<end>1690</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Mixed Distillation Helps Smaller Language Models Reason Better
%A Chenglin, Li
%A Chen, Qianglong
%A Li, Liangyue
%A Wang, Caiyu
%A Tao, Feng
%A Li, Yicheng
%A Chen, Zulong
%A Zhang, Yin
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Findings of the Association for Computational Linguistics: EMNLP 2024
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F chenglin-etal-2024-mixed
%X As large language models (LLMs) have demonstrated impressive multiple step-by-step reasoning capabilities in recent natural language processing (NLP) reasoning tasks, many studies are interested in distilling reasoning abilities into smaller language models (SLMs) via fine-tuning. Previous distillation methods usually utilize the capabilities of LLMs to generate chain-of-thought (CoT) samples to teach SLMs. However, this distillation approach performs poorly in certain scenarios due to the limitations of CoT. In this work, we introduce a novel Mixed Distillation (MD) framework, distilling multiple step-by-step reasoning abilities into SLMs. First, we leverage LLMs to generate multiple step-by-step reasoning rationales by sampling automatically. Then, we create high-quality, well-balanced mixed thought data and design a novel multi-task loss to help SLMs better learn and adaptively activate multiple step-by-step reasoning. Our extensive experiments demonstrate that MD enhances both single-path (using either CoT or PoT) and multi-path (using both CoT and PoT) reasoning abilities of SLMs during inference across reasoning tasks. Notably, a single model generated by MD exceeds the comprehensive performance of an ensemble of two individual CoT and PoT distilled models. Mistral-7B using MD can achieve remarkable improvements of 87.5%, 74.0% and 77.1% on SVAMP, GSM8K and ASDIV, respectively, outperforming the teacher model, GPT-3.5-Turbo. We hope our work provides insight into SLMs’ multiple step-by-step reasoning abilities.
%R 10.18653/v1/2024.findings-emnlp.91
%U https://aclanthology.org/2024.findings-emnlp.91/
%U https://doi.org/10.18653/v1/2024.findings-emnlp.91
%P 1673-1690
Markdown (Informal)
[Mixed Distillation Helps Smaller Language Models Reason Better](https://aclanthology.org/2024.findings-emnlp.91/) (Chenglin et al., Findings 2024)
ACL
- Li Chenglin, Qianglong Chen, Liangyue Li, Caiyu Wang, Feng Tao, Yicheng Li, Zulong Chen, and Yin Zhang. 2024. Mixed Distillation Helps Smaller Language Models Reason Better. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 1673–1690, Miami, Florida, USA. Association for Computational Linguistics.