FOCUS&RePAIR: Mitigating Text Degeneration via Token-Level Guidance for Pruned Large Language Models
A token-level analysis of this failure mode is presented by viewing decoding as a dynamical process that enters and persists in a small set of recurrent contexts and shows that persistence is controlled by the escape mass assigned to plausible alternatives within the token sampling set.
Junyoung Lee, Sehyeon Park, Shinhyoung Jang et al.
· 0 citations