Nice Fold or Hero Call: Learning Budget-Efficient Thinking under Policy-Dependent Solvability
Adaptive reasoning is formulated as a computational investment under uncertainty, where budget follows the expected return of reasoning rather than perceived difficulty, and Budget-Efficient Thinking (BET) is proposed, a two-stage framework that combines behavioral cold-start with GRPO under an investment-cost-aware re...
Zhaomeng Zhou, Lan Zhang, Jun-Yang Wang et al.
· 0 citations