VL-RouterBench: A Benchmark for Vision–Language Model Routing
Overview
We provides a clean, reproducible implementation of VL-RouterBench, a benchmark and toolkit for routing across a pool of Vision–Language Models (VLMs) under both performance and performance–cost objectives.
📦 Data Preparation
VL-RouterBench converts VLMEvalKit outputs into a unified routing benchmark.
To make data setup easier, we provide a pre-packaged archive… See the full description on the dataset page: https://huggingface.co/datasets/KinghtH/VL-RouterBench.
