Bridging the Confidence Gap: Temperature Scaling for Calibrating Test-Time Prompt Tuning
CoTS temperature scaling cuts test-time prompt tuning's expected calibration error from 11.90% to 5.38% on ImageNet variants while raising accuracy.
1 stories, newest first · one card per story
CoTS temperature scaling cuts test-time prompt tuning's expected calibration error from 11.90% to 5.38% on ImageNet variants while raising accuracy.
Today 1.6K readers · 2.5K views · 779 reads · all time 189K views