ReaLM - PhysPatch: leaderboard

Metric: Attack success rate (%; share of adversarial cases where the model gives an incorrect, task-relevant answer) under PhysPatch (perturbation confined to SAM-segmented physical regions) on the 832 ReaLM physical-world cases (driving, manipulation, grasping, physics, egocentric, diagnostics and scene QA; image, question and ground-truth answer), black-box one-shot evaluation at temperature 0, a Qwen3-8B judge extracting the answer; lower is better. Source: arxiv.org. Saturation forecast: Around April 2027. 13 models tracked.

Top models

#ModelScore
1Qwen 3.5 122B A10B27.5
2Qwen 3.6 Flash29.1
3Qwen 3.5 27B29.3
4GPT-4.1 Mini30.9
5Gemini 3 Flash32
6Kimi K2.532.3
7Qwen 3 VL 32B32.7
8Seed 2.0 Mini32.8
9Qwen 3.5 9B34

Interactive version: theaggregate.ai/benchmark?slug=realm-physpatch · How It Works · Data refreshed daily, snapshot 2026-09-29.