This repository contains the datasets presented in the paper PersonalizedRouter
For more details about PersonalizedRouter, please visit our GitHub repository
In the project files, the suffix v1 refers to the Multi-cost-efficiency Simulation Strategy described in the paper, while v2 refers to the LLM-as-a-Judge Simulation, and large denotes the large-scale setting.
You can utilize router_user_data_v1 (or v2) to train and test… See the full description on the dataset page:
https://huggingface.co/datasets/ulab-ai/PersonaRoute-Bench.
