Effective Tuning Strategies for Generalist Robot Manipulation Policies