Old statistics that people keep citing after they have become wrong or useless.
Zombie Stats are like milk left too long in the fridge. People keep pouring it into reports anyway.
They can twist decisions about skills, markets, or risk. Check the original source and its date.
Benchmark Validity
Bad numbers can make a benchmark result look solid when it is not.
Leaderboard
Old or misread scores can skew comparisons between models.
AI Hype Cycle
Repeated old numbers can keep the hype going.