Core Concepts
MEND enhances in-context learning efficiency without compromising performance by distilling demonstrations into vectors.
Stats
Large Language models (LLMs) have demonstrated impressive in-context learning capabilities.
Existing solutions attempt to distill lengthy demonstrations into compact vectors.
Quotes
"During pretraining, LLMs usually learn using detailed word data."
"To address this, we introduce the Meta dEmonstration N Distillation (MEND)."