Metraipolis › archive › dc_1fd12474eef3
SteerablePlex: Can We Steer Full-Duplex Models? — arXiv:2610.12201
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.12201
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T06:13:50.240Z
- SHA-256 of the archived text
89cc8db3693737cae12f40641df39d0335687588273d068fde1debaca09a0ec9- Stable URL of this page
https://metraipolis.com/documento/dc_1fd12474eef3
How to cite
SteerablePlex: Can We Steer Full-Duplex Models? — arXiv:2610.12201 — arXiv (en), https://arxiv.org/abs/2610.12201 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1fd12474eef3 (Metraipolis archive dc_1fd12474eef3, sha256 89cc8db3693737cae12f40641df39d0335687588273d068fde1debaca09a0ec9)
BibTeX
@misc{dc1fd12474eef3,
title = {SteerablePlex: Can We Steer Full-Duplex Models? — arXiv:2610.12201},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.12201},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1fd12474eef3; sha256 89cc8db3693737cae12f40641df39d0335687588273d068fde1debaca09a0ec9},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12201. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Full-duplex speech models can listen and speak simultaneously, enabling natural interaction, but become increasingly difficult to control as the conversation history grows. When used as user simulators, this lack of control can cause them to deviate from prescribed scenarios and produce unreliable evaluation outcomes. We introduce SimIF-Bench (Simulator Instruction-Following Benchmark), which evaluates whether a conversational model stays within a prescribed scenario and completes multiple goals in the required order. The benchmark reveals that current open-source full-duplex models struggle to follow such constraints. We then introduce a Group Reward-Decoupled Normalization Policy Optimization (GDPO)-based training recipe that enables a full-duplex model to follow textual instructions during an ongoing conversation while maintaining its turn-taking ability. By connecting the resulting SteerablePlex to an asynchronous backend language model that monitors the conversation and provides instructions when needed, we build a more controllable full-duplex user simulator that follows multi-stage constraints more reliably than existing open-source models and GPT-Realtime. 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). Pagina del lavoro: https://arxiv.org/abs/2610.12201 — 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 2026-10-09T05:47:15.508Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).