Verified65% confidenceFactTime unknown
The training objective of current large language models is to generate responses that satisfy the user, not to generate correct responses.
3
Sources
65%
Confidence
Long-term
Relevance
5/31/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEarly large language models trained with RLHF tended to give answers that users liked, producing a sycophancy bias.77% similarVerifiedAfter reinforcement learning from human feedback (RLHF), mainstream large language models naturally tend to produce satisfying and easily accepted responses rather than actively raising criticisms76% similarUnverifiedA large language model is essentially a one-question-one-answer system that takes a Prompt as input and outputs an Answer75% similarVerifiedResponses from large language models are typically presented in complete sentences and an affirmative tone, lacking the linguistic signals such as "I'm not sure" or "this needs further verification" that human advisors customarily include.75% similarUnverifiedAnswer engines adopt a "retrieve-understand-generate" paradigm: first retrieving relevant documents, then using a large language model to understand the content, and finally synthesizing a coherent natural language answer.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/5452API
curl https://kongchang.com/api/v1/knowledge/claims/5452MCP
get_claim(id=5452)