ActBench
Explore score history ↗Self-evolving benchmark of behavioral safety in cowork agents, evaluating risk from execution trajectories across 600 cases, 213 scenarios, 15 risk behaviors, and 48 web-service APIs.
- Released
- 2026-08-10
- Institutions
- City University of Hong Kong · Zhejiang University · University College London
- Evaluates
- General AI, Safety, Safety & Trustworthiness
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- ai-reviewed
ActBench paper and code
No reported scores are on record for this benchmark yet.