Training language models to follow instructions with human feedback
The InstructGPT paper applies demonstrations and preference feedback to align model outputs with user intent. It reports that a smaller aligned model can be preferred to a much larger base model on the studied prompts.
Overview. The InstructGPT paper applies demonstrations and preference feedback to align model outputs with user intent.
Context. It reports that a smaller aligned model can be preferred to a much larger base model on the studied prompts.
Claims and evidence
Claim ledger, 2The InstructGPT paper applies demonstrations and preference feedback to align model outputs with user intent.
It reports that a smaller aligned model can be preferred to a much larger base model on the studied prompts.