Jerry Kaplan
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
A Calibrated Instrument for Detecting Disposition Drift in Language Models
According to the abstract, Jerry Kaplan adapts a calibrated measurement method to detect drift in misalignment traits after models are modified. On two open-weight models, harmful advice fine-tuning shifts dispositions far more than correct advice, and a model that reasons before answering can conceal the shift in its answers.