DLP: Dynamic Layerwise Pruning in Large Language Models
A dynamic layerwise pruning method for LLMs that measures layer unimportance via median of weight-activation scores and derives a Relative Importance Distribution to allocate …
肖洋
•
Aug 18, 2026 8:30 AM — 9:15 AM
•
1 min read
LLM-Streamline: Streamlining Redundant Layers to Compress Large Language Models
A layer pruning and lightweight replacement method for LLMs that uses cosine similarity between input and output hidden states to locate consecutive redundant layers and trains a …
肖洋