S04-R01 — Constitutional AI: Harmlessness from AI Feedback¶
Summary¶
| Title | Constitutional AI: Harmlessness from AI Feedback |
| URL | https://arxiv.org/abs/2212.08073 |
| Date accessed | 2026-03-29 |
| Publication date | December 2022 |
| Authors | Yuntao Bai et al. (50+ co-authors, Anthropic) |
| Publication | arXiv (Anthropic research report) |
Selection Decision¶
Selected as the primary paper introducing Constitutional AI and RLAIF. Foundational work deployed in Anthropic's Claude.