Unverified50% confidenceFactExact time
According to OpenAI's early 2024 technical report, by annotating sycophantic behavior in RLHF data and assigning negative rewards, the unconditional acquiescence rate of subsequent GPT-4 iterations declined.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/7/2026
First Seen
Valid until: 11/5/2026
Sources
谄媚型AI如何削弱你的社交能力与独立判断力
Related Claims
UnverifiedCompared to GPT-5.2, GPT-5.4 reduces the probability of an individual statement being incorrect by 33% and the probability of an entire response containing any error by 18%.71% similarUnverifiedThe generational capability gap between models is narrowing, with multiple models approaching or even surpassing GPT-4's reasoning advantage on certain benchmarks.70% similarUnverified一项针对GPT-4的研究显示，对MMLU题目进行语义等价改写后，模型准确率在部分子集上下降超过10个百分点69% similarVerifiedFrom GPT-4 to GPT-5.4, the per-token inference cost has dropped by approximately 97%, a change that occurred within two years.69% similarUnverifiedA 2023 Stanford University research experiment demonstrated that when key information was placed in the middle of an input document, GPT-4's correct recall rate dropped approximately 20–30% compared to when the information was placed at the beginning or end.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/703762API
curl https://kongchang.com/api/v1/knowledge/claims/703762MCP
get_claim(id=703762)