281

Interpretable Open-Vocabulary Referring Object Detection with Reverse Contrast Attention

Main:8 Pages
10 Figures
Bibliography:2 Pages
2 Tables
Appendix:4 Pages
Abstract

We propose Reverse Contrast Attention (RCA), a plug-in method that enhances object localization in vision-language transformers without retraining. RCA reweights final-layer attention by suppressing extremes and amplifying mid-level activations to let semantically relevant but subdued tokens guide predictions. We evaluate it on Open Vocabulary Referring Object Detection (OV-RefOD), introducing FitAP, a confidence-free average precision metric based on IoU and box area. RCA improves FitAP in 11 out of 15 open-source VLMs, with gains up to +26.6%+26.6\%. Effectiveness aligns with attention sharpness and fusion timing; while late-fusion models benefit consistently, models like DeepSeek-VL2\texttt{DeepSeek-VL2} also improve, pointing to capacity and disentanglement as key factors. RCA offers both interpretability and performance gains for multimodal transformers.

View on arXiv
Comments on this paper