Confident in a Confidence Score: Investigating the Sensitivity of Confidence Scores to Supervised Fine-Tuning
It is found that calibration shifts substantially after SFT: across 216 configurations, it degrades in 112 cases and improves in 104, with confidence scores shifting due to factors beyond output quality, such as proximity to the training distribution.
Lorenzo Jaime Yu Flores, Cesare Spinoso di-Piano, Jackie C. K. Cheung
· 0 citations