AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

The Silent Thought: Modeling Internal Cognition in Full-Duplex Spoken Dialogue Models via Latent Reasoning

Donghang Wu, Tianyu Zhang, Yuxin Li, Hexin Liu, Chen Chen, Eng Siong Chng, Yoshua Bengio

ICML 2026 · Large language models › Language, safety and interpretability · approach: Speech and audio language models · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add latent reasoning during the listening phase (ablation)FLAIR · MMSUaccuracy: 50.2 -> 56.2Improved
add latent reasoning during the listening phase (ablation)FLAIR · Llama Questionsaccuracy: 73.0 -> 78.0Improved
add latent reasoning during the listening phase (ablation)FLAIR · OpenbookQAaccuracy: 72.9 -> 74.2Improved
add latent reasoning during the listening phase (ablation)FLAIR · SDQAaccuracy: 54.4 -> 56.2Improved
add latent reasoning during the listening phase (ablation)FLAIR · WebQuestionsaccuracy: 41.7 -> 43.0Improved
add latent reasoning during the listening phase (ablation)FLAIR · TriviaQAaccuracy: 53.8 -> 51.2Hurt
add latent reasoning during the listening phase (ablation)FLAIR · CommonEvalGPT score: 3.54 -> 3.65Improved
add latent reasoning during the listening phase (ablation)FLAIR · AlpacaEvalGPT score: 3.80 -> 3.85Improved
add latent reasoning during the listening phase (ablation)FLAIR · Impatient turn-taking latencylatency (s): 0.33 -> 0.39Hurt
add latent reasoning during the listening phase (ablation)FLAIR · Impatient barge-in latencylatency (s): 0.49 -> 0.46No clear effect
add latent reasoning during the listening phase (ablation)FLAIR · Full-Duplex-Bench barge-inGPT-4o score: 4.08 -> 4.22Improved
add latent reasoning during the listening phase (ablation)FLAIR · Full-Duplex-Bench barge-inTOR: 89.0 -> 92.0Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy