Comment by adamsmith143

Yeah man they have teams on standby to adjust the model whenever a random unknown author posts something on obscure pre-print servers. Then they spend hundreds of thousands of compute $ to improve the model on that one metric the paper attacks.