Fusing perceptual and dynamics anomaly scores enables online temperature scaling that cuts expected calibration error by 37% on physical DonkeyCar tests with four unseen anomaly types.
A baseline for detect- ing misclassified and out-of-distribution examples in neural networks,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Anomaly-Informed Confidence Calibration for Vision-Based Safety Prediction
Fusing perceptual and dynamics anomaly scores enables online temperature scaling that cuts expected calibration error by 37% on physical DonkeyCar tests with four unseen anomaly types.