self-supervised learning

Paper

Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

chain-of-thought: inductive-bias of data.

A Simple Contrastive Learning Objective for Alleviating Neural Text Degeneration

text degeneration