DiffuPlex: Accelerating Full-Duplex Spoken Dialog Models via Rolling Masked Diffusion

Key points

• We introduce DiffuPlex, a rolling masked diffusion framework that reduces this sequential computation by predicting multiple future user and assistant frames in a single backbone wake.1
• We consider two inference policies over the same predictor: DiffuPlex-LISTEN consumes multiple future frames when they predict assistant silence, whereas DiffuPlex-SPEAK can also consume predicted assistant speech.1
• Across full-duplex interaction and spoken-language evaluations, DiffuPlex substantially reduces sequential backbone computation while largely preserving interaction behavior and general capability.1
• DiffuPlex-LISTEN and DiffuPlex-SPEAK achieve $1.46\times$ and $1.59\times$ deployment-path wall-clock speedups and $1.61\times$ and $1.80\times$ Core LM speedups, with all measured backbone invocations completing within the 80ms interaction interval.1

Sources (1)

• [1]DiffuPlex: Accelerating Full-Duplex Spoken Dialog Models via Rolling Masked Diffusion
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 04:01 PM

  Recent full-duplex spoken dialog models enable simultaneous listening and speaking, but fine-grained models still advance their backbone autoregressively at every interaction frame.

  We introduce DiffuPlex, a rolling masked diffusion framework that reduces this sequential computation by predicting multiple future user and assistant frames in a single backbone wake.

Extractive summary: sentences quoted from the sources.