Unverified50% confidenceFactExact time
The attention mechanism of large language models may be drawn by the user's specific question, the names of political entities in the context, and political commentary in the training data, thereby probabilistically overpowering the influence of the system prompt.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedConversations with large language models can substantially influence users' attitudes and beliefs, spanning areas from product recommendations to political positions.77% similarUnverified长上下文模型在自我审查场景中因锚定效应可能带来更大偏差,早期生成内容对后续注意力计算影响越显著71% similarUnverifiedWhen users explicitly stated their own ideology, the model's responses shifted systematically toward the user's position.69% similarUnverifiedDifferent attention heads in multi-head attention tend to capture different types of semantic relationships, with some focusing on syntactic dependencies and others on coreference relations.68% similarUnverified长期依赖系统给出的建议消息可能削弱用户独立组织问题的能力,并形成提问方式的信息茧房68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1029202API
curl https://kongchang.com/api/v1/knowledge/claims/1029202MCP
get_claim(id=1029202)