Metraipolis › archive › dc_cdd941a9dbc8
Integrate the Reward, Don't Sample It: Window Likelihood for Interruptions and Backchannels in Full-Duplex Speech Models — arXiv:2610.35581
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.35581
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-27T08:13:22.257Z
- SHA-256 of the archived text
6eaeca79f035af6575fa3e82036acbfb48961d49858fdf7ba56c368c2012d389- Stable URL of this page
https://metraipolis.com/documento/dc_cdd941a9dbc8
How to cite
Integrate the Reward, Don't Sample It: Window Likelihood for Interruptions and Backchannels in Full-Duplex Speech Models — arXiv:2610.35581 — arXiv (en), https://arxiv.org/abs/2610.35581 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cdd941a9dbc8 (Metraipolis archive dc_cdd941a9dbc8, sha256 6eaeca79f035af6575fa3e82036acbfb48961d49858fdf7ba56c368c2012d389)
BibTeX
@misc{dccdd941a9dbc8,
title = {Integrate the Reward, Don't Sample It: Window Likelihood for Interruptions and Backchannels in Full-Duplex Speech Models — arXiv:2610.35581},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.35581},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cdd941a9dbc8; sha256 6eaeca79f035af6575fa3e82036acbfb48961d49858fdf7ba56c368c2012d389},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.35581. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A full-duplex spoken dialogue model must yield when the user cuts in and hold when the user merely backchannels. Sequence-level RL with a timing reward (GRPO) is the usual recipe; we test its whole-rollout form on four open models and find two failures. A binary window reward gives zero advantage on exactly the windows that most need learning (a zero-gradient trap), and the one score per rollout is spread over listening-state decisions the reward is not written in terms of (credit leakage), so the model talks into silence or stops answering. Both come from sampling a reward that is a known function of a few binary control decisions. We propose \emph{window likelihood} (\WL): sample a continuation per window, but integrate in closed form over the stop decisions it visits and maximise the log-probability of acting correctly---a noisy-OR for yielding, a survival product for holding. On HumDial-FDBench, \WL\ gives the best yield/hold F1 on all six structurally different open models we post-train, exceeds Kyutai's released GRPO checkpoints (trained on different data), and adds no re-opening of the floor on any model.
Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Audio and Speech Processing (eess.AS) ; Computation and Language (cs.CL); Sound (cs.SD).
Pagina del lavoro: https://arxiv.org/abs/2610.35581 — 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 2026-10-27T07:56:00.321Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).