Reproduce “Adaptive Knowledge Distillation for Device-Directed Speech Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hyung Gun Chi, Florian Pesce, Wonil Chang, Oggi Rudovic, Arturo Argueta, Stefan Braun, Vineet Garg, Ahmed Hussen Abdelaziz
The paper presents adaptive knowledge distillation for device-directed speech detection, training a small detector and task-specific heads on a frozen speech-recognition encoder together. The student learns from representations, frame-level attention, teacher predictions, and labels. In-house tests report average relative EER reductions of 26% for keyword invocations and 19% for follow-up conversations, but deployment latency and energy were not measured.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.