Unlocking Fine-Grained Translation Quality Estimation in LRMs through Mutually Boosting Implicit and Explicit Reasoning
This paper proposes a simple two-stage training framework that enables the mutually boosting of implicit (layer-wise) and explicit (token-wise) reasoning capabilities, and provides evidence for the mutually boosting between implicit and explicit reasoning.
R. Dang, X. Wang, Zhejian Lai et al.
· 0 citations