Constitutional AI vs. RLHF vs. Deliberative Alignment — LessWrong
lesswrong.com · 2,753 words · saved by 1 readers
Outline: • 1. Quick review of RLHF, Constitutional AI, and Deliberative Alignment for a somewhat-technical audience, literature review of historical…
x Constitutional AI vs. RLHF vs. Deliberative Alignment — LessWrong AI Alignment Fieldbuilding AI Control AI Frontpage 26 Constitutional AI vs. RLHF vs. Deliberative Alignment by laudiacay 11th Apr 2026 10 min read 0 26 Outline: Quick review of RLHF, Constitutional AI, and Deliberative Alignment for a somewhat-technical audience, literature review of historical failure modes. Introduce "Persona-Emotion-Behavior space"- combining two recent interpretability papers to get a loose framework for talking about personality stability and current alignment techniques What's going on with alignment in
saved by
related reading
- Alignment remains a hard, unsolved problem — LessWronglesswrong.com
- Constitutional AI: Harmlessness from AI Feedbackarxiv.org
- Current AIs seem pretty misaligned to me — LessWronglesswrong.com
- Constitutional AI: RLHF On Steroidsastralcodexten.substack.com
- Thoughts on the impact of RLHF research — LessWronglesswrong.com
- Self-Fulfilling Misalignment Data Might Be Poisoning Our AI Modelsturntrout.com
- Deep Deceptiveness — LessWronglesswrong.com
- Thoughts on the impact of RLHF research — AI Alignment Forumalignmentforum.org
- Constitutional AI: Harmlessness from AI feedbackanthropic.com
- [2605.10310] Positive Alignment: Artificial Intelligence for Human Flourishingarxiv.org
- Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment — LessWronglesswrong.com
- Should We Train Against (CoT) Monitors? — LessWronglesswrong.com