Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Cong, Quan, Xiaojun, Pan, Yan, Lin, Liang, Wu, Weigang, Chen, Xu
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2407.19807
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908398078918656
author Liu, Cong
Quan, Xiaojun
Pan, Yan
Lin, Liang
Wu, Weigang
Chen, Xu
author_facet Liu, Cong
Quan, Xiaojun
Pan, Yan
Lin, Liang
Wu, Weigang
Chen, Xu
contents We focus on the problem of fusing two or more heterogeneous large language models (LLMs) to leverage their complementary strengths. One of the challenges of model fusion is high computational load, specifically in fine-tuning or aligning vocabularies. To address this, we propose Cool-Fusion, a simple yet effective approach that fuses the knowledge of source LLMs, which does not require training. Unlike ensemble methods, Cool-Fusion is applicable to any set of source LLMs that have different vocabularies. To overcome the vocabulary discrepancies among LLMs, we ensemble LLMs on text level, allowing them to rerank the generated texts by each other with different granularities. Extensive experiments have been conducted across a variety of benchmark datasets. On GSM8K, Cool-Fusion increases accuracy from three strong source LLMs by a significant margin of 17.4\%.
format Preprint
id arxiv_https___arxiv_org_abs_2407_19807
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Cool-Fusion: Fuse Large Language Models without Training
Liu, Cong
Quan, Xiaojun
Pan, Yan
Lin, Liang
Wu, Weigang
Chen, Xu
Computation and Language
We focus on the problem of fusing two or more heterogeneous large language models (LLMs) to leverage their complementary strengths. One of the challenges of model fusion is high computational load, specifically in fine-tuning or aligning vocabularies. To address this, we propose Cool-Fusion, a simple yet effective approach that fuses the knowledge of source LLMs, which does not require training. Unlike ensemble methods, Cool-Fusion is applicable to any set of source LLMs that have different vocabularies. To overcome the vocabulary discrepancies among LLMs, we ensemble LLMs on text level, allowing them to rerank the generated texts by each other with different granularities. Extensive experiments have been conducted across a variety of benchmark datasets. On GSM8K, Cool-Fusion increases accuracy from three strong source LLMs by a significant margin of 17.4\%.
title Cool-Fusion: Fuse Large Language Models without Training
topic Computation and Language
url https://arxiv.org/abs/2407.19807