Unverified50% confidenceFactExact time
Recent research on representation engineering and activation steering has shown that certain semantic or stylistic attributes (such as sentiment倾向 and refusal behavior) have approximately linear directions in activation space, and adding or subtracting vectors along that direction can modulate model output.
1
Sources
50%
Confidence
Long-term
Relevance
10/28/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe output of large models has no emotion in essence, but factors such as anthropomorphic tendencies, trade-offs in instruction compliance, and in-context learning amplify users' perception of "temper."73% similarUnverifiedA user's negative tone in a conversation may trigger a more direct or blunt response style from the model through in-context learning mechanisms, rather than the model developing a long-term "personality."72% similarUnverifiedA model exhibiting avoidance behavior does not mean the model truly 'feels' pain; it is more likely a high-fidelity imitation of human behavioral patterns.70% similarUnverifiedThe PsyAgentBench study proposes that a model producing response patterns associated with a psychological effect and a model genuinely possessing that bias are two distinct propositions.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1108615API
curl https://kongchang.com/api/v1/knowledge/claims/1108615MCP
get_claim(id=1108615)