Liberating LLM Capabilities in Full-Duplex Speech Models
For developers of speech-based LLMs, LWS provides a method to liberate text-native capabilities (e.g., code generation, structured analysis) in realtime speech interaction, addressing a key limitation of current spoken-only models.
The paper proposes Listen-Write-Speak (LWS), a text-first tri-channel paradigm for speech-based LLMs that enables simultaneous listening, visible text writing, and spoken output. LWS achieves 92.6% writing-speaking consistency and strong performance on Full-Duplex-Bench and VoiceBench AlpacaEval (4.72), demonstrating that visible text can serve as a first-class output channel without sacrificing realtime responsiveness.
Speech-based large language models are typically constrained to spoken replies, which limits their user-facing outputs to what can be verbalized and suppresses text-native capabilities such as code generation, structured analysis, and multi-step reasoning in realtime interaction, for tasks that require persistent, structured, and inspectable intermediate outputs. Existing work improves spoken reasoning or full-duplex turn-taking, but still treats text as a hidden intermediate state or a subordinate modality rather than a first-class output channel. We propose Listen-Write-Speak (LWS), a text-first tri-channel paradigm in which a single autoregressive LLM continuously listens to user audio, writes visible free-form text as its primary output, and speaks a realtime oral response in parallel under a shared causal attention context. This behavior is implemented entirely through a Token Schema, requiring no architectural modifications, and learned via a two-stage data pipeline that synthesizes per-second cognitive annotations consistent with the revealed input timeline. Empirically, LWS demonstrates strong full-duplex interaction on Full-Duplex-Bench, reaches 4.72 on VoiceBench AlpacaEval, achieves 92.6% writing-speaking consistency, and consistently outperforms its internal ablations on URO-Bench. These results suggest that visible writing can serve as a first-class output channel for speech interaction without sacrificing realtime responsiveness. The code and dataset are available on the project page: https://royalzhang.com/project/lws-page/.