Same Beliefs, Different Policies: Learning Reusable Representations for Early Stopping in LLM Reasoning
BELT (BELief-based Termination), which separates outcome prediction from stopping control from end-to-end controllers, and its accuracy-cost trade-offs are competitive with or better than those of end-to-end controllers.
Jingwei Ni, Dmitry Abulkhanov, V. Smirnov et al.
· 0 citations