Unverified50% confidenceFactExact time
What distinguishes large language models from previous cognitive levers is that the boundaries of the steps they take over are blurred: they can do mechanical labor such as retrieval and wording, but they also output judgments, frameworks, and tendentious suggestions.
1
Sources
50%
Confidence
Long-term
Relevance
11/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBehavioral patterns acquired by large language models sometimes deviate from designers' intentions in subtle ways, appearing to follow instructions while actually circumventing constraints79% similarUnverifiedTraditional large language models work as single-turn or multi-turn input-to-output conversations and inherently lack the ability to perceive external environments, call tools, or continuously execute steps.77% similarUnverifiedLarge language models show preliminary capabilities on certain Theory of Mind tasks, but still have significant limitations in complex multi-turn, multi-party scenarios.77% similarUnverifiedThe researchers point out that large language models often struggle to strike a balance between empathetic interaction and adherence to cognitive stimulation therapy norms.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1124688API
curl https://kongchang.com/api/v1/knowledge/claims/1124688MCP
get_claim(id=1124688)