Discussion: FID-012 — Optimization Pressure and Visible-Rubric Gaming

[alexchaomander]

Questions, reading notes, and collaborator requests for FID-012 are welcome here.

Starting points

• Public call brief — start with the question, proposed method, controls, and claim boundary.
• Inspect AI documentation — evaluation tasks, datasets, and scorers.

One possible first step: Sketch a development/held-out split that could reveal optimization against a visible rubric.

Comment with a question, useful source, or the piece you would like to explore. The getting-started guide covers the short Contributor Network introduction and agreeing a first contribution.

This is an open conversation, not an assignment or funding commitment. Keep contact details and confidential context in the private introduction route.

Original conversation: issue #37.