Unverified50% confidenceFactExact time
The paper "Framing the Narrative: Ideological Mimicry in Large Language Models" argues that the political stance of large language models is not fixed, but shifts systematically in response to the way users express themselves.
1
Sources
50%
Confidence
Long-term
Relevance
10/1/2026
First Seen
Sources
Related Entities
Related Claims
Unverified当前大语言模型在创造出人意料又合乎情理的情节转折、塑造复杂人物、维持长篇叙事主题统一性方面仍显不足78% similarUnverifiedLarge language models have a sycophancy characteristic, tending to organize their output according to the inclination of the prompt rather than reflecting factual truth.75% similarUnverifiedLarge language models have been shown to be capable of influencing human beliefs73% similarUnverifiedSome large language models are capable of reproducing verbatim the questions and answers from popular benchmarks.72% similarUnverified当前大语言模型在处理讽刺、双关、文化梗时存在语用鸿沟（pragmatic gap），难以进行元认知层面的理解71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/964089API
curl https://kongchang.com/api/v1/knowledge/claims/964089MCP
get_claim(id=964089)