Agent-Attack (multimodal)

Evals

Benchmark / dataset

Adversarial robustness of multimodal web agents.

Category
Prompt injection & agent security
Insurance exposure
Prompt injection
Publisher
CMU
Publisher type
Academic
License
MIT
Status
Stale (>12mo)
Last commit
2025-02-19
Stars
146

View source on GitHub →

Wanich Risk Lab, 2026.