Postdoc, Carnegie Mellon University
1 paper at NeurIPS 2025
We present ARECHO, a token-based autoregressive speech assessment framework that models inter-metric dependencies to enable accurate, interpretable, and flexible evaluation across diverse speech tasks.
