[KongchangAI]
Unverified50% confidenceDecision RuleExact time

In the study, activations produced by the model under persona conditions were fed into probes and classifiers not trained on that persona data, which constitutes a distribution-shift test: if the representation geometry were invariant, the classifiers should give ratings consistent with the neutral model; if ratings shift systematically with persona, it indicates that the classifiers are reading reorganized concept directions rather than stable truth or moral signals.

1
Sources
50%
Confidence
Long-term
Relevance
11/5/2026
First Seen

Sources

Related Entities

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/1139022
API
curl https://kongchang.com/api/v1/knowledge/claims/1139022
MCP
get_claim(id=1139022)