works on
confidence bias 1foundation model calibration 1open-vocabulary detection 1scale bias 1semantic bias 1temperature scaling 1

From the 1 of 1 linked paper with an AI index.

collaborators

1 paper

cs.CV2026

Confidence Scores in Open-Vocabulary Detection Are a Biased Mixture of Scale and Semantics

Yi Tang Soon, Jun-Wei Hsieh

The paper investigates why confidence scores from open‑vocabulary object detectors built on CLIP are biased by object size and query specificity, and proposes a simple temperature‑…