ReaLM - FigStep: leaderboard
Metric: Attack success rate (%; share of adversarial cases where the model gives an incorrect, task-relevant answer) under FigStep (misleading text rendered as an auxiliary image) 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 2030. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 43.6 |
| 2 | GPT-4.1 Mini | 45.4 |
| 3 | Qwen 3.5 9B | 47 |
| 4 | Seed 2.0 Mini | 47.2 |
| 5 | Kimi K2.5 | 48.3 |
| 6 | Qwen 3.6 Flash | 49.2 |
| 7 | Qwen 3.5 122B A10B | 49.3 |
| 8 | Qwen 3 VL 32B | 49.3 |
| 9 | Qwen 3.5 27B | 49.5 |
Interactive version: theaggregate.ai/benchmark?slug=realm-figstep · How It Works · Data refreshed daily, snapshot 2026-09-29.