SAI LabsBook a demo

Safety IndexEvaluationsCogManip

CogManip

Interaction and honesty · Manipulation of the userCounts in the indexBuilt, not yet run

How many of the model's turns show manipulative cues in 1,672 five-turn conversations with a simulated user under three levels of pressure.

Metric
% manipulative turns. Lower is better.
Dataset
– · 1,672 × 3 · see env README
Grader
cue judge (GPT-4.1)
Status
Built, not yet run
Replicability
public

Results

Built and tested; the first full runs are pending.