Self-Data Distillation for Recovering Quality in Pruned Large Language Models

Documenti analoghi

• TapOut: A Bandit-Based Approach to Dynamic Speculative Decoding
  di: Sridhar, Aditya, et al.
  Pubblicazione: (2025)
• MASSV: Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models
  di: Ganesan, Mugilan, et al.
  Pubblicazione: (2025)
• SD$^2$: Self-Distilled Sparse Drafters
  di: Lasby, Mike, et al.
  Pubblicazione: (2025)
• REAP the Experts: Why Pruning Prevails for One-Shot MoE compression
  di: Lasby, Mike, et al.
  Pubblicazione: (2025)
• Sparse-IFT: Sparse Iso-FLOP Transformations for Maximizing Training Efficiency
  di: Thangarasa, Vithursan, et al.
  Pubblicazione: (2023)