Human-centered multimodal sexism detection for EXIST 2026 (CLEF) Task 2: a five-stream FiLM-conditioned cross-attention model that fuses meme text, image, annotator demographics, and biometric signals. Team VANGUARD.

Deep learning system for seismic velocity inversion using a SincNet-GAT-UNet architecture, physics-informed temporal encoding, graph attention fusion, and FiLM-conditioned spatial prediction

Extended MaxwellNet framework with modern deep learning architectures (FNO, Residual Attention UNet), feature-wise conditioning, and auxiliary electromagnetic regularization losses.