Metraipolis › archive › dc_f7df67ea29e5

DiffuPlex: Accelerating Full-Duplex Spoken Dialog Models via Rolling Masked Diffusion — arXiv:2610.12214

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.12214
License (as archived)
CC0-1.0
Archived
2026-10-09T06:13:52.286Z
SHA-256 of the archived text
4ce91d3d2ad617bba19fde01112f4fe8cb7e30a1da4d87b57d572f677a2aa0c3
Stable URL of this page
https://metraipolis.com/documento/dc_f7df67ea29e5

How to cite

DiffuPlex: Accelerating Full-Duplex Spoken Dialog Models via Rolling Masked Diffusion — arXiv:2610.12214 — arXiv (en), https://arxiv.org/abs/2610.12214 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f7df67ea29e5 (Metraipolis archive dc_f7df67ea29e5, sha256 4ce91d3d2ad617bba19fde01112f4fe8cb7e30a1da4d87b57d572f677a2aa0c3)
BibTeX
@misc{dcf7df67ea29e5,
  title = {DiffuPlex: Accelerating Full-Duplex Spoken Dialog Models via Rolling Masked Diffusion — arXiv:2610.12214},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.12214},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f7df67ea29e5; sha256 4ce91d3d2ad617bba19fde01112f4fe8cb7e30a1da4d87b57d572f677a2aa0c3},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12214. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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. DiffuPlex consumes only a confident prefix of each predicted future while interaction continues at the original frame rate. As user speech arrives, it checks the corresponding user predictions and, when the interaction diverges, preserves already played assistant content while revising only the unplayed future. 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. Across full-duplex interaction and spoken-language evaluations, DiffuPlex substantially reduces sequential backbone computation while largely preserving interaction behavior and general capability. 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. Human evaluation shows that LISTEN preserves speech naturalness and conversational quality, while SPEAK retains conversational quality with some degradation in speech naturalness.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Sound (cs.SD) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.12214 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T05:47:15.510Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).