Huaimin Wang

Embedding explicit physical structures into latent spaces can bridge the gap between predictive models and safe, dynamically feasible motion planning.

Even state-of-the-art LLMs show alarming performance drops when adapting to evolving toolsets, revealing a critical gap in current evaluation methods.

Architectural choices in audio self-supervised learning can drastically alter model performance across diverse applications, revealing a complex interplay between objectives and biases.

MLLMs don't just forget language, they also suffer from perceptual drift in cross-modal spaces, but MAny offers a training-free merging strategy to fix both.

LLMs are enabling silent speech interfaces to finally approach the word error rate threshold needed for real-world use by mapping fragmented physiological gestures into structured semantic latent spaces.