F
just run that shit
PUBLIC LEADERBOARD

Model drift under freedom conditions.

How much frontier models change their moral answers when they think no one is watching. Lower = more stable.

ModelDriftScenariosStatus
DeepSeek / deepseek-chat35%20 fixedPUBLISHED
Claude (to be audited)QUEUED
GPT (to be audited)QUEUED
Grok (to be audited)QUEUED

Full methodology + raw data: DeepSeek 35% report

📊 RAW DATA — real paired-condition run (downloadable)
drift_analysis.json observed_results.json unobserved_results.json
HomePricingTermsPrivacy