Preventing Catastrophic Overfitting in Fast Adversarial Training
Discover new insights and strategies to prevent catastrophic overfitting in fast adversarial training for more robust neural networks.
SycoPhantasy: Measuring Sycophancy in Small Vision-Language Models
Explore how SycoPhantasy quantifies sycophancy and hallucination in small VLMs assessing fantasy character image-text alignment.
Enhancing VLM Reasoning with Visual Cues & Reflection
Discover how ForeSight boosts Vision-Language Models' reasoning using low-level visual cues and mask-based feedback for superior AI understanding.
X-NegoBox: Secure Privacy Budgeting for P2P Energy Data
Discover X-NegoBox, a framework enabling explainable, adaptive privacy budgeting for secure peer-to-peer energy data exchange and enhanced prosumer trust.
Self-Abstraction Learning for Stable Deep Neural Training
Discover how Self-Abstraction Learning improves stability and effectiveness in training deep neural networks by addressing common optimization challenges.