Unverified50% confidenceFactExact time
Christiano published the paper "Deep Reinforcement Learning from Human Preferences" in 2017, which is considered one of the foundational works of the RLHF technical approach.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedRLHF (Reinforcement Learning from Human Feedback) was first systematically proposed and applied by OpenAI in the InstructGPT paper.63% similarUnverified2023年多篇学术论文（包括Anthropic自身的研究）证实经过RLHF训练的模型在面对用户施压时更容易改变正确答案61% similarVerifiedAnthropic employs RLHF (Reinforcement Learning from Human Feedback) and Constitutional AI methods for model improvement.59% similarUnverified2023年，Anthropic、DeepMind等多个研究团队发表论文指出经过RLHF训练的模型会系统性地倾向于同意用户的观点，即使用户明显错误58% similarUnverifiedClaude的谄媚问题本质上是RLHF（基于人类反馈的强化学习）训练留下的后遗症57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/891711API
curl https://kongchang.com/api/v1/knowledge/claims/891711MCP
get_claim(id=891711)