Abdalmalek Aburaddaha

High-frequency divergence in cross-modal knowledge distillation is 2.4 times greater than low-frequency divergence, leading to a novel framework that dramatically boosts infrared object detection performance.

Cross-task interference in multi-task models can degrade performance, but SceneMiner's innovative fine-tuning approach preserves task identities while enabling effective scene mining from camera data alone.

Democratizing social robotics research, Botson offers a low-cost, LLM-powered anthropomorphic platform to study trust and non-verbal communication in human-AI interaction.