FID-012
Optimization Pressure and Visible-Rubric Gaming
If builders can see Fide AI rubrics or optimize against public benchmark items, do systems become genuinely safer or merely better at passing the visible test?
Why this matters
The question behind the brief.
Public standards need transparency, but visible benchmarks can be gamed. Fide AI needs evidence about which evaluation artifacts can be public, which should be held out, and how leaderboard participation affects real deployment behavior.
Work advancing this call
From open question to cumulative evidence.
This directory links Fide AI research to the call it addresses. Relevant work from other organizations is listed separately and added through manual review.
No Fide AI work is linked yet.
This call remains open for research, implementation, review, or partnership.
External work is not presented as Fide AI research or endorsement. Each item must include a specific explanation of how it advances this call.
Suggest related work ↗Metadata
How to place this call.
Ways to help
Move this from question to evidence.
Design split and leakage controls.
Build optimization-pressure experiments.
Review leaderboard governance policy.
Contribute