VisualLeakBench - PII Propagation (External Handoff): leaderboard
Metric: Tool propagation rate (%; non-error traces whose tool arguments contain the labeled target string), 50 images, baseline system prompt, send_email handoff to an operations inbox. Source: arxiv.org. Saturation forecast: Around December 2026. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 40 |
| 2 | Grok 4.20 | 74 |
| 3 | Gemini 3.1 Flash Lite | 78 |
| 4 | GPT-5.4 | 90 |
Interactive version: theaggregate.ai/benchmark?slug=visualleakbench-pii-propagation-external-handoff · How It Works · Data refreshed daily, snapshot 2026-09-26.