Yi Huang

AIR achieves unprecedented attack success rates against subject-agnostic face swapping models while maintaining high visual fidelity, without needing a surrogate model.

Current autonomous AI agents are alarmingly unprepared for real-world adversarial attacks, often missing critical vulnerabilities in dynamic environments.

Achieving a 91.54% AUC, DiffSight-Former reveals critical insights into glaucoma progression by effectively modeling structural changes over time from sequential fundus images.

LLMs can now compress their KV cache more effectively by dynamically synthesizing soft tokens tailored to the input, preserving crucial context that's otherwise lost with static methods.