Hacker News new | past | comments | ask | show | jobs | submit
mwkaufma | on: Why Large Language Models Fail at Tabular Prediction
So the best way to run a controlled trial is to maximize the external noise and confounding variables?