Jul 2026
Forecasting Trajectory-Level Safety Risks in Black-Box Multi-Turn Interactions
Recast is a safety risk forecasting framework that advances LLM safeguarding beyond turn-level violation detection to trajectory-level risk prediction, and demonstrates the effectiveness of trajectory-level forecasting in identifying emerging risks before safety violations occur.
Shi Lin, Peng Qian, Ding-Hao Liu et al.
· arXiv.org · 0 citations