Reference

This section documents the evaluation reference material: config structure, CLI usage, supported models, and how to read outputs.

Page	What you get
Configuration	Hydra YAML layout under evaluations/<benchmark>/, including runner, env, and rollout fields required for embodied_eval.
CLI	How to launch evaluations with run_eval.sh, pass Hydra overrides, and auto-infer the benchmark from config names.
Models	VLA models with example configs in evaluations/ today (OpenPI, OpenVLA-OFT, StarVLA, DreamZero, LingBotVLA) and how to set model_path.
Results	Where logs and rollout videos are written, terminal metrics such as eval/success_once, and TensorBoard usage.