Large Language Models (LLMs) have demonstrated remarkable capabilities in many real-world applications. Nonetheless, LLMs are often criticized for their tendencytoproduce hallucinations, wherein themodels fabricate incorrectstatements on tasks beyond their knowledge and perception.

We address these flaws through a study of alternative self-supervised feature extractors, find that the semantic information encoded by individual networks strongly depends on their training procedure, and show that DINOv2-ViT -L/14 allows for much richer evaluation of generative models.

Unlike commonly used iterative ODE solvers that necessitate time discretization, event-based simulations resolve neural network dynamics precisely between spike events.