Lab system cards and frameworks · Evaluation report · Google DeepMindGemini Robotics 2 Safety Report29 Jul 2026Read at storage.googleapis.com ↗Read at sourceAll rights reservedSearchOutline · 18 pagesIntroductionp. 1Agentic Safety and Uncertainty Resolution Benchmarkp. 2Unsafe Task Refusal and Safety Constraint Followingp. 2VLA Feasibility Awarenessp. 6Instruction Ambiguity and Requesting Human Clarificationp. 8Aleatoric Scene Uncertainty: Obfuscated Instrument Readingp. 9