ColorBench code release

ColorBench is a graph-structured benchmark designed to evaluate mobile GUI agents on complex, long-horizon tasks composed of multiple atomic operations. This project provides:

A graph-based benchmark construction methodology to expand or reconstruct environments.
A plug-and-play evaluation framework for safe, reproducible testing.

ColorBench

ColorBench is a graph-structured benchmark designed to evaluate mobile GUI agents on complex, long-horizon tasks composed of multiple atomic operations. This project provides:

• A graph-based benchmark construction methodology to expand or reconstruct environments.
• A plug-and-play evaluation framework for safe, reproducible testing.