Conformal prediction on vision-language models for action recognition reduces class-set sizes, but the proposed calibration-set temperature tuning violates the claimed 1-alpha coverage guarantee.
Human-in-the-loop vehicle reid,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Conformal Predictions for Human Action Recognition with Vision-Language Models
Conformal prediction on vision-language models for action recognition reduces class-set sizes, but the proposed calibration-set temperature tuning violates the claimed 1-alpha coverage guarantee.