Your fact-checks

Your fact-checks will appear here

factually
Support us
  • Home
  • Politics
  • Business
  • Society
  • Technology
  • Health
  • Science
  • Entertainment
Index/Topics/RLHF

RLHF

Constitutional AI uses AI-generated supervision instead of large volumes of human preference labels, similar to RLHF.

Fact-Checks

1 result
Feb 8, 2026

How does Anthropic’s Constitutional AI work and what audits exist for Claude?

’s "" trains Claude by embedding an explicit, public "constitution" of prioritized principles that the model uses to critique and revise its own outputs during fine-tuning, replacing much human gradin...

About
Blog
Contact
FAQ
Terms & ConditionsTerms
Privacy PolicyPrivacy
Manage data