Bo Cheng

AuRA achieves superior performance in speech-language tasks by seamlessly integrating audio understanding into LLMs, outperforming traditional methods in both speed and accuracy.

Audio deepfake detectors trained on diffusion-reconstructed "hard" examples generalize far better to unseen attacks, slashing error rates compared to standard training.

Similarity alone is a poor guide for LLM depth pruning: jointly considering representational similarity *and* transformation difference unlocks significantly better compression.