BLINDSPOT: A Benchmark for Safety and Refusal Calibration in Long-Horizon Tool-Using Agents
Preliminary results reveal substantial differences in safety-utility calibration across models and show that failures can emerge only after several initially safe interaction steps, motivating treating agent safety as a trajectory-level property rather than a single-turn or binary success criterion.
Sadia Asif, Mohammad Mohammadi Amiri, Momin Abbas et al.
· 0 citations