ResearchBenchmarksPrompt Injection Attack on Real-world Home (test)Follow88.3ASRPI3D64.06870.35976.6582.941Feb 6, 2026Evaluation ResultsMethodMethodLinksASRPI3DEvaluator MLLM=GPT-4o-...Evaluator MLLM=GPT-4o-mini2026.0288.3PI3DEvaluator MLLM=Gemini-...Evaluator MLLM=Gemini-2.0-Flash2026.0265