Lei Yu

Hallucination Reduction with CASAL: Contrastive Activation Steering for Amortized Learning

VL-JEPA: Joint Embedding Predictive Architecture for Vision-language

Misaligned Roles, Misplaced Images: Structural Input Perturbations Expose Multimodal Alignment Blind Spots

Emergence of a High-Dimensional Abstraction Phase in Language Transformers

Robust LLM safeguarding via refusal feature adversarial training

Intrinsic Evaluation of Unlearning Using Parametric Knowledge Traces

GEOMETRIC SIGNATURES OF COMPOSITIONALITY ACROSS A LANGUAGE MODEL’S LIFETIME