zstanjj/ClawTrojan download history
zstanjj/ClawTrojan is a text classification dataset on the Hugging Face Hub. In the last 30 days it was downloaded 9,921 times (111 in the last 7 days), and 20,848 times in total. It ranks #3,045 among datasets by monthly downloads.
ClawTrojan ClawTrojan is a long-horizon agent safety evaluation dataset for studying multi-stage, stealthy attack trajectories against tool-using agents. Instead of modeling a sample as a single malicious prompt, ClawTrojan models an attack as a trajectory that unfolds across user req
Spaces using ClawTrojan
- Model Pulse 8 likes