weaker

The paradox of LLM self-distillation: Faster reasoning, weaker generalization – TechTalks

The Rise of Self-Distillation in LLM Optimization The quest for more efficient and performant large language models has driven significant…