Calibrating Vision-Language Models for Out-of-Distribution Detection
IEEE Signal Processing Letters · 2026 · 33 · 4136-4140
DOI: 10.1109/lsp.2026.3736348 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Qichao Chen und 5 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Gefördert von: Key Research and Development Program of Guangdong Province.
Abstract (Originalsprache)
Vision–language models (VLMs) have shown strong potential for out-of-distribution (OOD) detection by matching test images with textual prototypes derived from in-distribution (ID) class names. However, generic class prompts mainly encode object identity and leave the dataset-level visual context implicit, which may cause semantic underspecification in CLIP-based OOD scoring. We address this issue from the perspective of Domain Calibration and propose Domain-Context Prompting (DCP), a zero-shot framework that calibrates the full set of ID textual prototypes with shared domain semantics. DCP obtains a dataset-level domain anchor through an LLM-assisted candidate generation and OOD-free semantic consensus selection procedure, and injects the selected anchor into class-conditioned prompts. The resulting OOD score is still computed by maximum cosine similarity over ID classes, but each prototype encodes joint class-and-domain information. Under additional few-shot ID supervision, the same calibrated prototype space can be optionally refined with lightweight Low-Rank Adaptation (LoRA), yielding DCP+. Experiments on standard ImageNet-1 K OOD benchmarks demonstrate the effectiveness of DCP in the zero-shot setting and further improvements from its few-shot extension.
Förderung
- Key Research and Development Program of Guangdong Province
Zitieren
APA
Chen, Q., Wu, Y., Chen, Z., Maul, T., Zhou, D., & Wen, J. Z. (2026). Calibrating Vision-Language Models for Out-of-Distribution Detection. IEEE Signal Processing Letters, 33, 4136-4140. https://doi.org/10.1109/lsp.2026.3736348
Vancouver
Chen Q, Wu Y, Chen Z, Maul T, Zhou D, Wen JZ. Calibrating Vision-Language Models for Out-of-Distribution Detection. IEEE Signal Processing Letters. 2026;33:4136-4140. doi:10.1109/lsp.2026.3736348
BibTeX
@article{Chen2026,
author = {Chen, Qichao and Wu, Yanqi and Chen, Zhiyuan and Maul, Tomas and Zhou, Decheng and Wen, James Zhiqing},
title = {Calibrating Vision-Language Models for Out-of-Distribution Detection},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {4136--4140},
doi = {10.1109/lsp.2026.3736348}
}
EUVIMED