PARE-Bench

Benchmark-oriented Python research framework for evaluating proactive AI assistants through active user simulation; the official docs center on benchmark sweeps, scenario generation, and annotation workflows.