Unverified60% confidenceFactExact time
Constitutional AI is divided in engineering implementation into a supervised learning phase (SL-CAI) and a reinforcement learning phase (RL-CAI), where the RL phase uses the AI's own preference judgments as reward signals and optimizes the policy model via the PPO algorithm.
2
Sources
60%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
Anthropic发起生命科学黑客松，10万美元赋能AI科研
twitterclaudeai7/2/2026
Related Claims
VerifiedCurrent mainstream AI alignment methods include RLHF, Constitutional AI, and DPO, which internalize safety constraints into model weights during the model training phase.77% similarVerified宪法AI通过两个阶段完成训练：第一阶段让模型根据宪法原则对自身输出进行批判和修订，第二阶段利用AI反馈替代部分人工标注来训练偏好模型74% similarVerifiedConstitutional AI is a training method proposed by Anthropic in 2022, whose core idea is to provide an AI system with a set of explicit written principles and introduce an AI feedback (RLAIF) mechanism74% similarVerifiedConstitutional AI enables models to internalize norms through self-criticism and iterative correction based on preset value principles, without requiring large-scale human preference annotation data, distinguishing it from RLHF.73% similarUnverified过去打造AI Agent主要依赖强化学习（RL）算法，需要为每一个任务训练一个专门的模型72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/162508API
curl https://kongchang.com/api/v1/knowledge/claims/162508MCP
get_claim(id=162508)