Foyer study: logistic regression beats LLMs on inbox routing
On about 200 labels, logistic regression led every field, while the most accurate system ignored 18% of emails needing action.
3 stories, newest first · one card per story
On about 200 labels, logistic regression led every field, while the most accurate system ignored 18% of emails needing action.
SLITE uses 17 linguistic features for textual entailment, reaching 83% on SICK and nearing RoBERTa.
Argues LLM classifiers should feed a downstream logistic regression, yielding calibration, threshold control, and principled use of structured covariates.
Today 1.6K readers · 2.4K views · 775 reads · all time 189K views