MetaLLM

This repository contains the code for the paper MetaLLM: A High-performant and Cost-efficient Dynamic Framework for Wrapping LLMs:

• benchmark: Code to evaluate the performance of base LLMs,
• get_embedding_mmlu: Code to get the embedding to train the router,
• online_router: Train and evaluate the MetaLLM.

@article{nguyen2024metallm,
  title={MetaLLM: A High-performant and Cost-efficient Dynamic Framework for Wrapping LLMs},
  author={Nguyen, Quang H and Dao, Thinh and Hoang, Duy C and Decugis, Juliette and Manchanda, Saurav and Chawla, Nitesh V and Doan, Khoa D},
  journal={arXiv preprint arXiv:2407.10834},
  year={2024}
}