Verbosity bias is the one I'd watch. A longer estimate isn't a better estimate, and a judge that rewards length will push everything toward padding. Checking the judge against a few human-graded answers is the step we won't skip at ML Systems.
Agreed on verbosity bias. One cheap extra check on your human-graded set: correlate the judge's scores with answer length. If that correlation is stronger than the one in the human scores, the bias is showing. The post suggests aiming for roughly 0.85 correlation with human raters and re-checking regularly.
Salman Parvez | ML Systems
Verbosity bias is the one I'd watch. A longer estimate isn't a better estimate, and a judge that rewards length will push everything toward padding. Checking the judge against a few human-graded answers is the step we won't skip at ML Systems.