Fetching the paper…

Generalization v.s. Memorization: Tracing Language Models' Capabilities Back to Pretraining Data · Around