A survey proposing an L0-L3 architectural hierarchy, T×I×R interaction ontology, and IDLE/LISTEN/SPEAK/WAIT/DUAL decision state machine for full-duplex spoken dialogue systems, documenting a realization gap between architectural potential and observed behavior due to training data limits.
hub Canonical reference
In2024 IEEE Spo- ken Language Technology Workshop (SLT), pages 1115–1122
Canonical reference. 80% of citing Pith papers cite this work as background.
26
Pith papers citing it
Background
80% of classified citations
abstract
Spoken dialogue modeling poses challenges beyond text-based language modeling, requiring real-time interaction, turn-taking, and backchanneling. While most Spoken Dialogue Models (SDMs) operate in half-duplex mode-processing one turn at a time - emerging full-duplex SDMs can listen and speak simultaneously, enabling more natural conversations. However, current evaluations remain limited, focusing mainly on turn-based metrics or coarse corpus-level analyses. To address this, we introduce Full-Duplex-Bench, a benchmark that systematically evaluates key interactive behaviors: pause handling, backchanneling, turn-taking, and interruption management. Our framework uses automatic metrics for consistent, reproducible assessment and provides a fair, fast evaluation setup. By releasing our benchmark and code, we aim to advance spoken dialogue modeling and foster the development of more natural and engaging SDMs.
hub tools
citation-role summary
background 4
other 1
citation-polarity summary
representative citing papers
FD-SLMs exhibit state inertia during abrupt interruptions that a training-free perception-vector steering intervention mitigates, lifting correctness from 28% to 45% and IWOR from 40% to 72% on the Zero-Buffer Benchmark.
SpeechEditBench provides seven atomic editing tasks, compositional multi-operation instructions, and an anchor-based protocol yielding target success, preservation success, and joint success metrics; evaluations show no model excels across dimensions and compositional editing is especially difficult
Introduces the first full-duplex audio-visual-to-audio-visual benchmark, showing current agents fail to use visual nonverbal cues in live conversation.
DuplexSLA introduces a three-channel full-duplex architecture that synchronizes continuous user audio, discrete assistant audio, and rate-limited textual actions inside a single backbone for native turn-taking and in-conversation tool use.
EVA-Bench supplies a simulation engine for bot-to-bot voice dialogues plus two composite metrics (EVA-A for accuracy, EVA-X for experience) evaluated on 213 enterprise scenarios, showing no tested system exceeds 0.5 on both pass@1 scores.
TiCo enables spoken dialogue models to follow explicit time constraints in generated responses using Spoken Time Markers and reinforcement learning with verifiable rewards, cutting duration error by 2.7x over its backbone.
Causal full-duplex speech LLMs improve response quality by recursively feeding soft vocabulary embeddings as latent thoughts during listening, trained by matching a non-causal expert posterior via ELBO.
Game-Time Benchmark shows spoken language models handle basic tasks but degrade sharply under temporal constraints like tempo adherence and synchronized responses.
Gemini 2.5 Flash matches human rank correlation within 0.07 on 5 of 8 dimensions and within-1 agreement on 6 of 8 when judging raw stereo full-duplex agent conversations, with rank order transferring across later Gemini models.
Lychee-FD resolves modality interference in full-duplex spoken language models by separating acoustic and semantic parameters in deep layers and adding a dense semantic alignment channel, achieving state-of-the-art performance on spoken QA and interaction benchmarks.
A causal future voice-activity model’s negative log-likelihood, pooled over turn-taking boundary units, ranks natural dyadic clips above five classes of controlled timing perturbations.
FacePlex introduces a unified streaming model with Rolling Flow Matching and Rolling Cross-Attention to enable full-duplex joint real-time generation of speech and facial motion tokens.
A multi-axis RL alignment technique improves pause handling, turn-taking, backchanneling, and interruption response in full-duplex spoken dialogue models by optimizing axis-specific rewards derived from human audio segments.
Omni-DuplexEval provides a new benchmark and automatic evaluation method for real-time duplex omni-modal interaction, showing state-of-the-art models reach only 39.6% overall and 20% on proactive reminders.
Engagement Process (EP) decouples actions and observations as independent event streams over time within a POMDP structure to explicitly model temporal dynamics in agent interactions.
ASPIRin decouples speaking timing from token content via binary action space projection and applies GRPO with rule-based rewards to optimize interactivity in SLMs without semantic collapse or repetition.
MM-tau-p² is a new benchmark with 12 metrics that measures how well multi-modal agents adapt to user personas and maintain robustness in dual-control interactions.
The survey introduces a four-category taxonomy for LALM evaluations and reviews benchmarks across general auditory processing, knowledge reasoning, dialogue, and fairness-safety.
LLMs beat humans and supervised models at next speaker prediction in meetings using only text, while multimodal LLMs improve on addressee and turn-change tasks but remain below human performance.
A speech-based model forecasts conversation turn endpoints up to 2.56 seconds ahead to enable lower-latency spoken dialogue via speculative LLM and TTS execution.
Full-duplex SDMs show strong representational synchronization that peaks near zero lag and degrades with noise, with internal states encoding anticipatory turn-taking cues detectable ahead of time.
A survey of Large Audio Language Models that establishes a taxonomy of trustworthiness vulnerabilities and proposes a Defense-in-Depth roadmap for audio intelligence.
A new HumDial-FDBench benchmark and real human-recorded dual-channel dataset are released to assess full-duplex dialogue systems on interruptions and conversational flow.
citing papers explorer
-
A Survey of Full-Duplex Spoken Dialogue Systems: Architectural Hierarchy, Interaction Ontology, and Decision State Machine
A survey proposing an L0-L3 architectural hierarchy, T×I×R interaction ontology, and IDLE/LISTEN/SPEAK/WAIT/DUAL decision state machine for full-duplex spoken dialogue systems, documenting a realization gap between architectural potential and observed behavior due to training data limits.
-
Overcoming State Inertia in Full-Duplex Spoken Language Models via Activation Steering
FD-SLMs exhibit state inertia during abrupt interruptions that a training-free perception-vector steering intervention mitigates, lifting correctness from 28% to 45% and IWOR from 40% to 72% on the Zero-Buffer Benchmark.
-
SpeechEditBench: A Bilingual Multi-Attribute Benchmark for Instruction-Guided Speech Editing
SpeechEditBench provides seven atomic editing tasks, compositional multi-operation instructions, and an anchor-based protocol yielding target success, preservation success, and joint success metrics; evaluations show no model excels across dimensions and compositional editing is especially difficult
-
VideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents
Introduces the first full-duplex audio-visual-to-audio-visual benchmark, showing current agents fail to use visual nonverbal cues in live conversation.
-
DuplexSLA: A Full-Duplex Spoken Language Model with Synchronized Speech, Language, and Action
DuplexSLA introduces a three-channel full-duplex architecture that synchronizes continuous user audio, discrete assistant audio, and rate-limited textual actions inside a single backbone for native turn-taking and in-conversation tool use.
-
EVA-Bench: A New End-to-end Framework for Evaluating Voice Agents
EVA-Bench supplies a simulation engine for bot-to-bot voice dialogues plus two composite metrics (EVA-A for accuracy, EVA-X for experience) evaluated on 213 enterprise scenarios, showing no tested system exceeds 0.5 on both pass@1 scores.
-
TiCo: Time-Controllable Spoken Dialogue Model
TiCo enables spoken dialogue models to follow explicit time constraints in generated responses using Spoken Time Markers and reinforcement learning with verifiable rewards, cutting duration error by 2.7x over its backbone.
-
The Silent Thought: Modeling Internal Cognition in Full-Duplex Spoken Dialogue Models via Latent Reasoning
Causal full-duplex speech LLMs improve response quality by recursively feeding soft vocabulary embeddings as latent thoughts during listening, trained by matching a non-causal expert posterior via ELBO.
-
Game-Time: Evaluating Temporal Dynamics in Spoken Language Models
Game-Time Benchmark shows spoken language models handle basic tasks but degrade sharply under temporal constraints like tempo adherence and synchronized responses.
-
A Reliability Assessment of LALM Audio Judges for Full-Duplex Voice Agents
Gemini 2.5 Flash matches human rank correlation within 0.07 on 5 of 8 dimensions and within-1 agreement on 6 of 8 when judging raw stereo full-duplex agent conversations, with rank order transferring across later Gemini models.
-
Hierarchical Acoustic-Semantic Modeling: Modality Separation and Semantic Coherence for Full-Duplex SLMs
Lychee-FD resolves modality interference in full-duplex spoken language models by separating acoustic and semantic parameters in deep layers and adding a dense semantic alignment channel, achieving state-of-the-art performance on spoken QA and interaction benchmarks.
-
TurnNat: Automatic Evaluation of Turn-Taking Naturalness in Dyadic Spoken Dialogue
A causal future voice-activity model’s negative log-likelihood, pooled over turn-taking boundary units, ranks natural dyadic clips above five classes of controlled timing perturbations.
-
FacePlex: Full-Duplex Joint Speech-Facial Motion Generation for Conversational Avatars
FacePlex introduces a unified streaming model with Rolling Flow Matching and Rolling Cross-Attention to enable full-duplex joint real-time generation of speech and facial motion tokens.
-
Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models
A multi-axis RL alignment technique improves pause handling, turn-taking, backchanneling, and interruption response in full-duplex spoken dialogue models by optimizing axis-specific rewards derived from human audio segments.
-
Omni-DuplexEval: Evaluating Real-time Duplex Omni-modal Interaction
Omni-DuplexEval provides a new benchmark and automatic evaluation method for real-time duplex omni-modal interaction, showing state-of-the-art models reach only 39.6% overall and 20% on proactive reminders.
-
Engagement Process: Rethinking the Temporal Interface of Action and Observation
Engagement Process (EP) decouples actions and observations as independent event streams over time within a POMDP structure to explicitly model temporal dynamics in agent interactions.
-
ASPIRin: Action Space Projection for Interactivity-Optimized Reinforcement Learning in Full-Duplex Speech Language Models
ASPIRin decouples speaking timing from token content via binary action space projection and applies GRPO with rule-based rewards to optimize interactivity in SLMs without semantic collapse or repetition.
-
MM-tau-p$^2$: Persona-Adaptive Prompting for Robust Multi-Modal Agent Evaluation in Dual-Control Settings
MM-tau-p² is a new benchmark with 12 metrics that measures how well multi-modal agents adapt to user personas and maintain robustness in dual-control interactions.
-
Towards Holistic Evaluation of Large Audio-Language Models: A Comprehensive Survey
The survey introduces a four-category taxonomy for LALM evaluations and reviews benchmarks across general auditory processing, knowledge reasoning, dialogue, and fairness-safety.
-
Evaluating Large Language Models Abilities for Addressee, Turn-change, and Next Speaker Prediction in Meetings
LLMs beat humans and supervised models at next speaker prediction in meetings using only text, while multimodal LLMs improve on addressee and turn-change tasks but remain below human performance.
-
Endpoint Anticipation for Low-Latency Spoken Dialogue
A speech-based model forecasts conversation turn endpoints up to 2.56 seconds ahead to enable lower-latency spoken dialogue via speculative LLM and TTS execution.
-
Synchronization and Turn-Taking in Full-Duplex Speech Dialogue Models
Full-duplex SDMs show strong representational synchronization that peaks near zero lag and degrades with noise, with internal states encoding anticipatory turn-taking cues detectable ahead of time.
-
A Survey of Large Audio Language Models: Generalization, Trustworthiness, and Outlook
A survey of Large Audio Language Models that establishes a taxonomy of trustworthiness vulnerabilities and proposes a Defense-in-Depth roadmap for audio intelligence.
-
Full-Duplex Interaction in Spoken Dialogue Systems: A Comprehensive Study from the ICASSP 2026 HumDial Challenge
A new HumDial-FDBench benchmark and real human-recorded dual-channel dataset are released to assess full-duplex dialogue systems on interruptions and conversational flow.
-
FastTurn: Unifying Acoustic and Streaming Semantic Cues for Low-Latency and Robust Turn Detection
FastTurn fuses streaming CTC semantics with acoustic cues for lower-latency, more accurate turn detection in full-duplex dialogue and releases a real-dialogue test set.
-
Toward Native Multimodal Modeling: A Roadmap
A roadmap that defines architectural nativity for multimodal models and categorizes them into Multi-to-Text, Multi-to-Target, and Multi-to-Multi types while outlining an industrial pipeline toward unified transformer-based native multimodal modeling.