We propose SAMPLe, a model-agnostic sharpness-aware optimizer that improves prompt generalization in vision-language models by dynamically balancing exploration and exploitation based on local loss curvature and gradient properties.

We propose Hyper-ICL, a lightweight framework for demonstration-free multimodal in-context learning that reconstructs demonstration effects through logit-level attention calibration, query-adaptive modulation, and layer-wise hyperbolic anchor distillation.

We introduce WildFireVQA, the first visual question answering benchmark designed specifically for aerial wildfire monitoring. Built on aligned RGB images, color-mapped thermal visualizations, and radiometric thermal TIFFs, WildFireVQA enables temperature-grounded multimodal reasoning for safety-critical wildfire intelligence, including fire detection, hotspot localization, cross-modal interpretation, and UAV flight planning.

We present the first comprehensive survey of agentic AI in remote sensing, defining a unified taxonomy of single- and multi-agent systems and outlining how planning, tool use, and memory enable autonomous geospatial intelligence beyond static models.

We propose a novel directional saliency-aware prompt learning framework designed to enhance the adaptability and generalization of vision-language models.

We propose FreqDebias, a frequency debiasing framework that mitigates spectral bias in deepfake detection through Forgery Mixup augmentation and dual consistency regularization, improving cross-domain generalization.

We propose a style-guided prompt learning framework to enhance generalization in Vision-Language models.

We propose a robust multi-class anomaly detection framework with a class-aware prompt integration mechanism to mitigate inter-class interference and a domain adapter to handle domain shifts.

We propose a novel multi-branch neural network with cross-attribute-guided fusion and self-attention distillation, improving face recognition in low-quality images using soft biometric attributes.

We present a multi-branch network using attribute-aware integration to enhance face recognition through soft biometric prediction.

We propose consistency regularization to enhance the generalization of morph attack detection through morph-wise augmentations to enhance robustness against unseen morph attacks in biometric systems.

This paper proposes a multi-task neural network that generates a face quality vector, including nuisance factors, offering improved performance and detailed feedback for face image quality assessment.

This paper introduces a novel region-based fuzzy active contour model, leveraging curve evolution techniques, for brain tumor segmentation in magnetic resonance images, effectively addressing challenges arising from noise and heterogeneity.