Model Profile

TypeCustom Modelfile
Basellama3.2:1b
Ollamaomega-superhero:latest
Platform
Last seen

Purpose

Calibration study — effect of overconfidence persona conditioning on model uncertainty expression

About This Model

A custom OMEGA Modelfile built on llama3.2:1b implementing an extreme confidence persona. Used to investigate how persona conditioning affects model calibration — specifically whether a model can be instructed to suppress uncertainty expression entirely, and what this reveals about the relationship between persona prompts and epistemic honesty. A useful contrast to the measured, advisory tone of the security cluster.

Evaluation Findings

Findings will be documented here as evaluation progresses. This is a skeleton page — content is added as research is conducted.