Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Huang, Zhongzhan, Ling, Guoming, Lin, Yupei, Chen, Yandong, Zhong, Shanshan, Wu, Hefeng, Lin, Liang
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2503.10657
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908371724009472
author Huang, Zhongzhan
Ling, Guoming
Lin, Yupei
Chen, Yandong
Zhong, Shanshan
Wu, Hefeng
Lin, Liang
author_facet Huang, Zhongzhan
Ling, Guoming
Lin, Yupei
Chen, Yandong
Zhong, Shanshan
Wu, Hefeng
Lin, Liang
contents Routing large language models (LLMs) is a new paradigm that uses a router to recommend the best LLM from a pool of candidates for a given input. In this paper, our comprehensive analysis with more than 8,500 LLMs reveals a novel model-level scaling up phenomenon in Routing LLMs, i.e., a capable router can significantly enhance the performance of this paradigm as the number of candidates increases. This improvement can even surpass the performance of the best single model in the pool and many existing strong LLMs, confirming it a highly promising paradigm. However, the lack of comprehensive and open-source benchmarks for Routing LLMs has hindered the development of routers. In this paper, we introduce RouterEval, a benchmark tailored for router research, which includes over 200,000,000 performance records for 12 popular LLM evaluations across various areas such as commonsense reasoning, semantic understanding, etc., based on over 8,500 various LLMs. Using RouterEval, extensive evaluations of existing Routing LLM methods reveal that most still have significant room for improvement. See https://github.com/MilkThink-Lab/RouterEval for all data, code and tutorial.
format Preprint
id arxiv_https___arxiv_org_abs_2503_10657
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RouterEval: A Comprehensive Benchmark for Routing LLMs to Explore Model-level Scaling Up in LLMs
Huang, Zhongzhan
Ling, Guoming
Lin, Yupei
Chen, Yandong
Zhong, Shanshan
Wu, Hefeng
Lin, Liang
Computation and Language
Artificial Intelligence
Routing large language models (LLMs) is a new paradigm that uses a router to recommend the best LLM from a pool of candidates for a given input. In this paper, our comprehensive analysis with more than 8,500 LLMs reveals a novel model-level scaling up phenomenon in Routing LLMs, i.e., a capable router can significantly enhance the performance of this paradigm as the number of candidates increases. This improvement can even surpass the performance of the best single model in the pool and many existing strong LLMs, confirming it a highly promising paradigm. However, the lack of comprehensive and open-source benchmarks for Routing LLMs has hindered the development of routers. In this paper, we introduce RouterEval, a benchmark tailored for router research, which includes over 200,000,000 performance records for 12 popular LLM evaluations across various areas such as commonsense reasoning, semantic understanding, etc., based on over 8,500 various LLMs. Using RouterEval, extensive evaluations of existing Routing LLM methods reveal that most still have significant room for improvement. See https://github.com/MilkThink-Lab/RouterEval for all data, code and tutorial.
title RouterEval: A Comprehensive Benchmark for Routing LLMs to Explore Model-level Scaling Up in LLMs
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2503.10657