ForeSight: Enhancing Risk Monitoring via Early Safety Signal Distillation
This work proposes ForeSight, a first-token output-risk forecasting framework that distills weak and redundant early safety signals into compact, layer-aware risk representations and achieves superior and efficient early-risk forecasting while relying solely on first-token hidden states.
Han-Lin Wang, Chen-Long Wei, Ling Xu et al.
· 0 citations