Unverified75% confidenceFactTime unknown
OpenAI通过RLHF（基于人类反馈的强化学习）和专项安全微调使模型能够识别自杀意念、自我伤害等高风险内容模式
1
Sources
75%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOpenAI's InstructGPT paper demonstrates that a 1.3B-parameter model trained with RLHF outperforms an unaligned 175B model in human evaluations.69% similarUnverifiedModern large model alignment training relies heavily on high-quality preference data pairs, with OpenAI's RLHF and RLAIF being representative examples68% similarUnverifiedOpenAI extensively used 'instruction following' as a reward signal in ChatGPT's RLHF training process, causing the model to tend toward executing user requirements one by one.67% similarUnverifiedCompanies such as OpenAI and Anthropic have publicly acknowledged the side effect of sycophantic tendencies arising from RLHF training.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/18433API
curl https://kongchang.com/api/v1/knowledge/claims/18433MCP
get_claim(id=18433)