Imitation Learning vs Trial-and-Error Learning

Definition

Karpathy's framing distinguishing two types of learning: imitation (pre-training, supervised fine-tuning — watch and repeat) versus trial-and-error reinforcement learning (learning from outcomes). The latter produces emergent, surprising strategies because the model's cognition differs from the human labeler's.

Cast within

No topic-region cast yet — this appears once Imitation Learning vs Trial-and-Error Learning's compiled claims are aligned into a topic region's argument tree.

Claims mentioning Imitation Learning vs Trial-and-Error Learning (2)

Your notes

Loading notes...