Daily Papers — 2026-09-11
27 papers on audio, speech, music, and acoustics.
1. CMA-OT: Hierarchical Expert Supervision for Dance-to-Music Generation
Authors: Jinting Wang, Chenxing Li, Dong Yu, Li Liu
Categories: cs.AI, cs.SD | 12 pages, 9 figures, accepted by ACM MM 2026
Dance-to-music (D2M) generation aims to synthesize music that is rhythmically and stylistically aligned with dance videos. A key challenge arises from the semantic mismatch between sparse dance cues, such as rhythm and style, and the dense information required for music composition, including structure, instrumentation, and expressive dynamics. Existing methods typically rely on these sparse cues and supervise only the final audio output, resulting in poorly learned music representations and generated music with limited musicality and structural coherence. To address these issues, we propose Curriculum-guided Multi-scale representation Alignment with scale-aware Optimal Transport (CMA-OT), a novel paradigm that leverages an external music expert to provide hierarchical supervision for the generator’s latent features, bridging the semantic gap and enhancing representation learning. To effectively incorporate hierarchical supervision, we introduce a curriculum-guided multi-scale learning strategy that progressively transfers musical knowledge from the expert to the music generator, enabling stable and effective representation learning. Moreover, to accommodate the semantic and structural variations across different expert scales and achieve fine-grained alignment under temporal mismatch, we propose a scale-aware optimal transport alignment mechanism, which models soft correspondences between hierarchical expert representations and the generator’s latent features. Extensive experiments on two datasets demonstrate that CMA-OT achieves state-of-the-art performance in rhythmic synchronization, perceptual quality, and overall music generation.
2. Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents
Authors: Yunqi Lu, Tyler Baumgartner, Nikhil Johri, Brandon Tai, Candice Fan et al.
Categories: cs.CL, cs.SD | 12 pages, 8 tables
Full-duplex evaluation often emphasizes whether an agent keeps speaking or stops. That binary cannot express a third response humans use routinely: continuing to speak while incorporating what the listener just contributed. The contribution may be a missing word, a correction or a clarification. We introduce Duplex Cue, an evaluation of this \emph{in-turn adaptation} in full-duplex voice agents. Duplex Cue separates listener intent (backchannel, collaboration, or interruption) from speaker behavior: continuing unchanged, adapting within the turn, or yielding. Adaptation includes acknowledgment as well as content revision. In a single-model case study using 300 human-confirmed cues from unscripted English conversations, we compare recorded human responses with PersonaPlex continuations generated while replaying the listener’s audio. We retain 208 pairs with the ongoing speaker active at cue onset and a scorable response in each condition. On the 66 collaborative pairs, recorded speakers adapt in 68.2\% of cases, compared with 34.8\% for PersonaPlex. The model otherwise continues unchanged (42.4\%) or yields (22.7\%). These findings show why evaluating natural voice interaction requires measuring how an agent responds to a listener’s contribution as well as whether it keeps speaking.
3. Kraken: LLM-based Speech-to-Speech Translation via Low-bitrate VQ and Dual-path Source Conditioning
Authors: Hayato Futami, Hassan Shahmohammadi, Tushar Dhyani, Alkis Koudounas, Raphaël Lafargue et al.
Categories: cs.CL, cs.SD
Speech-to-speech translation (S2ST) has advanced significantly with speech LLMs, offering the potential for joint optimization and preserving non-linguistic information. However, these models struggle with predicting high-bitrate speech tokens in LLMs, and face the challenge of relying on S2ST training data with ideally aligned speaker identity and prosody. We propose using low-bitrate tokens based on single-layer vector quantization, trained to reconstruct self-supervised learning (SSL) features. We also employ a separate token-to-waveform decoder named Autowave-X, which is also conditioned on the source speech to improve non-linguistic transfer, thereby relaxing the training data constraints. With the integration of these techniques, we propose an S2ST model named Kraken, which augments a pre-trained LLM with speech feature inputs and the low-bitrate token outputs, followed by Autowave-X vocoder. We built the model upon Qwen3-8B and trained it using 150k hours of multilingual and multitask speech data. We demonstrated that our model exhibited better translation quality than SeamlessM4T-Large v2 and Qwen2.5-Omni, along with improved speaker and prosody transfer capabilities.
4. CVSS-X: A Multilingual Speech-to-Speech Translation Corpus for 28 Languages
Authors: Lucas Rafael Stefanel Gris, Alef Iury Siqueira Ferreira, Frederico Santos de Oliveira, Augusto Seben da Rosa, Alexandre Costa Ferro Filho et al.
Categories: cs.CL, cs.SD, eess.AS | Accepted at the SALMA Workshop (2nd Edition) @ EMNLP 2026 (Non-archival)
We introduce CVSS-X, a large-scale synthetic speech-to-speech translation corpus that extends CVSS by reversing the translation direction. While CVSS translates from 21 languages into English, CVSS-X enables translation from English into 28 target languages spanning 12 language families. The corpus comprises approximately 240,000 parallel speech pairs per language, totaling over 16,000 hours, eight times larger than CVSS. We provide two variants: CVSS-X-C with two canonical voices per language, and CVSS-X-T with cross-lingual voice cloning, both fully generated. Evaluation shows comparable translation quality to CVSS with consistent performance across typologically diverse languages. Combined with CVSS, this enables research on bidirectional and multilingual speech-to-speech translation. The code is available at https://github.com/ErmisAI/XVSS-X and the dataset under CC-BY-NC 4.0 license at https://huggingface.co/datasets/lgris/XVSS-X.
5. TokenMapper: A Step Toward Interoperable Speech Token Translation
Authors: Tal Kozakov, Tal Rosenwein, Eliya Nachmani
Categories: cs.LG, cs.AI, cs.SD, eess.AS | Accepted to AACL-IJCNLP 2026 findings
Neural audio codecs discretize speech into token sequences, but the resulting token spaces differ in vocabulary and codebook structure, preventing direct communication across models. This limitation affects applications such as conversational voice agents and speech to speech translation systems where multiple speech models must interact. As a result, transferring information between speech systems typically requires decoding to waveform audio and re-encoding with a second tokenizer, increasing latency and introducing potential information loss. To address these limitations, we present TokenMapper, a direction aware framework for direct token to token translation between heterogeneous speech tokenizers in the discrete domain. TokenMapper supports structurally mismatched token spaces, including mappings between single codebook and multi codebook representations, under a shared effective token rate. Experiments on GLM-4-Voice, MiMi and DualCodec show consistent cross model performance. Specifically, translation WER approaches native reconstructions within 2.5-6.8% absolute WER, human MOS for TokenMapper outputs ranges from 2.29 to 4.39, following the same direction level trends as UTMOS and end to end latency is reduced by 4.8-94.5% relative to waveform bridging, reaching up to 972 ms per utterance. These results provide a practical step toward cross model speech token interoperability without intermediate waveform reconstruction.
6. DiffSynth-Music: Audio-Conditioned KV-Cache Adapters for Controllable Music Generation
Authors: Zhongjie Duan, Shengchuan Gao, Hong Zhang, Yingda Chen
Categories: cs.SD | 10 pages, 1 figure
Text and lyrics specify broad musical characteristics and sung content but offer limited control over musical timing, melody, and reference-based style. We introduce DiffSynth-Music (https://modelscope.cn/models/DiffSynth-Studio/DiffSynth-Music), a framework that adds composable audio conditioning to a music synthesis backbone through layer-wise key-value injection. The three template models, Control, Prosody, and Reference, are initialized from the backbone diffusion transformer and trained with conditional flow matching. They support five control types: beats, vocals, accompaniment, prosody, and reference audio. A shared variational autoencoder maps conditioning waveforms into a common latent space, enabling their attention memories to be combined. With the template timestep fixed at the clean-data endpoint and other inputs held constant, each control cache is computed once and reused throughout sampling. Training pairs are derived from music recordings using beat extraction, source separation, vocal resynthesis, and reference-excerpt selection. Single-control evaluations on Mandarin and English songs demonstrate improved adherence across all five control types and better lyric fidelity under vocal conditioning relative to the backbone. Automatic music-quality and instruction-following scores remain broadly comparable to those of the evaluated base models, with metric-specific trade-offs. We release the three template models to support research and creative applications in controllable music generation.
7. PhaseGAN: High-Fidelity Vocoder via Decoupled Amplitude and GAN-Driven Phase Reconstruction
Authors: Wenzheng Zhang, Xueliang Zhang, Shulin He, Fei Zhao, Xin Liu et al.
Categories: cs.SD | 15 pages, 1 figure
A vocoder is a pivotal component of modern text-to-speech (TTS) systems. Despite the significant progress of neural network-based vocoders, accurate phase reconstruction remains the main challenge limiting both audio quality and modeling efficiency. We introduce PhaseGAN, a lightweight vocoder that addresses this limitation through a “mel $\rightarrow$ Amplitude $\rightarrow$ Phase” reconstruction pipeline. By reconstructing amplitude and phase spectra via distinct methodologies, the proposed PhaseGAN outperforms state-of-the-art baselines while utilizing fewer model parameters and reduced computational requirements. The compact version generates high-fidelity audio with approximately 500K parameters and 1 GMAC computational load, making it highly suitable for real-time applications on edge devices. In addition, our approach exhibits exceptional musical audio synthesis capabilities despite no training on musical data, illustrating unprecedented cross-domain generalization. See https://github.com/phasegan/phasegan-audio-demo for demos of our work.
8. Direct Preference Density Alignment for Conversational Audio Equalization
Authors: Ioannis Stylianou, Sven Ewan Shepstone, Jon Francombe, Pablo Martinez Nuevo, Zheng-Hua Tan
Categories: cs.SD, cs.AI
Large Language Model alignment typically relies on learned proxy reward models, which significantly increase the memory footprint during training and are notoriously prone to instability and reward hacking. While offline methods like Direct Preference Optimization (DPO) bypass the reward model, they lose the ability to perform online exploration. If no optimization constraints are applied, this can lead to format collapse in bounded, continuous spaces. To resolve this, we propose Direct Preference Density Alignment: An alternative framework that removes the need for a learned proxy reward model while strictly preserving the benefits of online reinforcement learning. We leverage large-scale user data (approximately 90,000 samples) to construct non-parametric preference density maps, establishing an empirical reward surface. In addition to removing the reward model, Direct Preference Density Alignment enables the combination of the online structural grounding of Group Relative Policy Optimization (GRPO) with the targeted offline refinement of DPO. We show that this GRPO+DPO combination achieves the highest performance, and in a blind audio equalization listening test, enables a 1.5B-parameter model to achieve perceptual parity with a carefully prompt-engineered GPT-4o mini baseline, using only a fraction of the inference compute.
9. Predictive audio representations for early detection and tracking of hidden dynamic objects
Authors: Katerina Vinciguerra, Moritz Brandes, Danilo Hollosi, Letizia Marchegiani
Categories: cs.SD, cs.AI, eess.AS | 8 pages, 2 tables, 6 figures
Predicting potential dangers is core to safety. Forecasting the presence of other traffic agents is core to danger prediction. Occluded traffic agents challenge detection systems as they might become visible too late, leaving the autonomous vehicle too little time to identify, plan and act accordingly in a robust and safe way. Previous works proved that auditory perception, being omnidirectional and not constrained by a field-of-view, provides fundamental cues for early spotting of different road users, even when hidden by other vehicles or infrastructures. Yet, those contributions deal with scenarios with only one vehicle present, and they either identify the type of vehicle or estimate its direction of arrival. In this work, we move forward, and propose a multi-task system which, simultaneously, estimates the number of vehicles present, their type, and their direction of arrival. Our methodological contribution is a two-stage pipeline: a self-supervised pre-training stage inspired by the Joint- Embedding Predictive Architecture (JEPA) applied directly to multichannel raw waveforms, followed by supervised multi-task fine-tuning with a bidirectional LSTM and three classification heads. The pre-training stage trains the encoder without labels, pushing it to predict the latent representation of a future audio segment from its past context. To train and test our framework, since no suitable dataset was publicly available, we collected an ad-hoc one covering Non-Line-Of-Sight scenarios with multiple traffic agents simultaneously operating. Experimental evaluation shows that our method outperforms the state of the art; it also proves that our design choice allows the model to learn robust representations, which can be transferred to an unseen driving scenario, maintaining reasonable and stable performance.
10. What Did the MLLM Hear? Token-Level Spectro-Temporal Grounding for Audio MLLM Explainability
Authors: Lucia Cascone, Valeria Fraenza, Michele Nappi, Fabio Narducci, Benedetto Simone
Categories: cs.SD, cs.CV, cs.LG
Audio-based Multimodal Large Language Models (MLLMs) can generate detailed natural-language descriptions of complex acoustic scenes, yet it remains unclear which parts of the input audio support each generated token. This is particularly challenging because acoustic evidence is distributed across time and frequency, and concurrent sound events may overlap temporally while occupying different spectral regions. We introduce STAG, to our knowledge the first post-hoc framework for token-level spectro-temporal grounding of captions generated by audio-based MLLMs. STAG estimates the temporal support for each generated token using target-token-specific vocabulary projections of the encoded audio representations, measures frequency-band relevance through controlled spectral occlusion, and combines the two signals into a spectro-temporal relevance map. We evaluate STAG against ten post-hoc explanation methods across four grounding benchmarks, where it achieves the best event-localization performance on every dataset, and apply it to eight audio-language backbones without parameter updates. Counterfactual deletion further shows that removing the identified evidence selectively reduces confidence in the corresponding event and frequently removes it from the regenerated caption. These results provide behavioral support for the faithfulness and selectivity of the explanations.
11. StepAudio 3 Gen Technical Report
Authors: Bin Lin, Bo Zhao, Boyang Wang, Boyang Zhang, Boyong Wu et al.
Categories: cs.SD, eess.AS
We introduce StepAudio 3 Gen, a general-purpose audio generation model that supports zero-shot text-to-speech (TTS), voice design, vocal generation, sound effects, music, vibe speech, and mixtures of multiple audio types within a unified framework. At its core, StepAudio 3 Gen is a discrete autoregressive generator that models audio directly over residual vector quantization (RVQ) tokens, departing from the diffusion Transformer-based continuous generation paradigm prevalent in recent general audio models. Its StepAudio Tokenizer represents general audio at 12.5 Hz in a shared $16 \times 2048$ residual code space, jointly quantizing semantic and waveform-level acoustic features so that each code layer preserves both types of information. For generation, the backbone predicts the first codebook along the time axis using autoregressive modeling, while a lightweight causal Transformer completes the remaining fifteen codebooks along the codebook axis. Our study further identifies three key design principles: (1) interference-aware progressive pretraining for acquiring audio capabilities while preserving the textual abilities of the large language model, (2) RVQ Adaptor for effectively incorporating multi-codebook acoustic representations, and (3) discrete autoregressive modeling over a shared representation across general audio domains. With progressive pretraining, multi-task instruction training, and supervised fine-tuning, StepAudio 3 Gen achieves state-of-the-art performance on both TTS and voice design, while retaining strong generation capabilities across speech, vocals, sound effects, and music. Audio samples are available at https://stepaudiollm.github.io/step-audio-3-gen/.
12. A Device to Control and Manipulate Occlusion Effects for Own Voice Perception Studies
Authors: Rouben Rehman, Simon Kersten, Aron Schliep, Janina Fels
Categories: eess.AS
The occlusion effect (OE) refers to changes of the eardrum sound pressure through ear canal occlusion. It consists of two phenomena: an insertion loss (IL) attenuating air-conducted sounds, and an occlusion gain (OG) amplifying bone-conduction. Perceptual research on this is hindered by high variability of the OE across individuals, complicating repeatable presentation of precise OE conditions. Consequently, a method to control OE conditions reproducibly during perceptual experiments is needed. For such investigations, the system must enable separate control of IL and OG. We present an approach based on modified commercial earmuffs with integrated microphones. The design integrates dedicated impedance measurements and the derivation of digital filters, which are applied to the microphone signals to emulate arbitrary OE curves. The system is evaluated objectively through appropriate measurements. Results show that the headphones’ inherent OG is guaranteed to be below 6dB above 115Hz, falling below 0dB above 155Hz. Emulation is shown to work accurately over the entire frequency range of interest. Limitations arise mainly due to a slight residual inherent OG for deep voices and the processing delay of the system. Future work will focus on the perceptual evaluation of the system and its application in perceptual studies.
13. AlignDPO: Preference-Gated Alignment for Reducing Hallucination in Decoder-Only TTS
Authors: Xiao Zhou, Oisín Turbitt, Kit Bower-Morris, Jonathan Carlton, Jamie Stacey et al.
Categories: eess.AS | 6 pages, 4 figures, 4 tables. Accepted to IEEE SLT 2026
Decoder-only text-to-speech (TTS) models scale efficiently but remain prone to content hallucinations that arise from weak text-speech alignment during autoregressive generation. We find that robustness is governed by a non-monotone relation to the sharpness of the alignment-bearing attention heads: a moderate degree is best, whereas over-sharpening is no better than the unaligned backbone and even less robust. Guided by this, we present AlignDPO, a post-training method that reaches this moderate regime by folding a lightweight connectionist-temporal-classification (CTC) alignment term into Direct Preference Optimization (DPO), applied only to the chosen samples, with no architectural or inference-time change. On the Seed-TTS-Eval English set, this significantly reduces the content-hallucination and word error rates relative to a strong DPO baseline and lowers the severe content-hallucination rate to ~0.6% (from 4.4%); a listening study further finds it preferred for naturalness over both the backbone and that baseline. Alignment is thus best learned and kept moderate rather than maximized or imposed at decoding. Audio samples are available at https://align-dpo-demo.vercel.app.
14. Location-based Training with Complementary Folded Linear Orderings for Multichannel Speech Separation
Authors: Kaixuan Yang, Stijn Kindt, Nilesh Madhu
Categories: eess.AS | Accepted by IWAENC 2026
Location-based training (LBT) effectively resolves the output permutation problem in multichannel speech separation by imposing deterministic spatial orderings. For planar microphone arrays, LBT typically adopts circular azimuth ordering to cover the full spatial range. However, the resulting cyclic topology introduces a discontinuity at the wrap-around point, increasing learning complexity and limiting the effective use of spatial cues. This work investigates this limitation by introducing location-based training with folded linear orderings (LBT-FLOs), which collapse circular azimuths into controlled linear orderings. While individual LBT-FLOs exhibit front-back ambiguity, each provides enhanced spatial discriminability over specific azimuth regions. Exploiting their complementarity, we propose an ensemble-style framework that selects among multiple LBT-FLOs using azimuth-guided scoring. Experiments across planar array geometries and reverberant conditions show modest but consistent improvements over circular-ordering LBT, with robustness to azimuth estimation errors.
15. Objective Intelligibility Prediction Using Distance Metrics on Speech Foundation Model Representations
Authors: Lyonel Behringer, Andreas Brendel
Categories: eess.AS | (c) 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
High-dimensional representations of pretrained speech foundation models have proven beneficial for objective speech quality and intelligibility prediction. While existing work on neural intelligibility prediction usually leverages such representations for task-specific fine-tuning, in this work we evaluate the usefulness of such representations for intelligibility prediction without any further training. We conduct a layer-wise analysis of multiple speech foundation models, correlating various embedding distances with subjective intelligibility scores. The results show that embeddings extracted from Whisper speech recognition models are best suited, with the last encoder and decoder layers yielding the best correlations when using the Fréchet Audio Distance. Notably, the evaluated distances outperform classical intelligibility metrics and are more robust than Word and Character Error Rates. Further, correlations improve with increasing size of the Whisper model from which embeddings are extracted.
16. VoxTubeS: Distributable Speaker-Anonymized Synthetic Speech Corpora and Their Analysis
Authors: Zhe Zhang, Yexin Lu, Junichi Yamagishi
Categories: eess.AS
Large speech corpora support research, but recordings can expose speaker identity because voice remains a recognizable biometric. Meanwhile, speech data derived from media can be difficult to redistribute reliably. We present \emph{VoxTubeS}, a family of speaker-anonymized synthetic speech corpora designed for redistribution, comprising three method families and seven variants derived from the VoxTube corpus, which is distributed under CC BY-NC-SA 4.0, using 1.29M quality-filtered English utterances from 1,511 speakers. The synthesis methods span voice conversion, latent-space anonymization, and controllable text-to-speech. We evaluate VoxTubeS using utterance-level unlinkability, conversation-level linkability and singling-out, downstream speaker verification, linguistic consistency, speaker diversity, and fairness metrics for gender and accents. Our comprehensive analysis exposes a complex trade-off: stronger identity suppression often reduces linkability but sacrifices utility and population diversity, whereas speaker consistency training improves both utterance- and conversation-level privacy while retaining comparable utility and a broader speaker space. Fairness varies independently of aggregate performance. No method dominates; VoxTubeS therefore treats corpus construction as a choice among operating points that balances privacy, utility, diversity, fairness, and responsible redistribution under the source license.
17. X-Pred MeanFlow for Streaming Token-to-Mel Speech Decoding
Authors: Hanke Xie, Xiaming Ren, Qirui Zhan, Jingbin Hu, Wenhao Li et al.
Categories: eess.AS | Accepted By ISCSLP2026
Recent advancements in discrete token-based speech generation have highlighted the importance of efficient token-to-waveform synthesis in streaming and dialogue scenarios. Flow-matching acoustic decoders achieve high-quality token-to-mel generation, but their iterative sampling requires multiple neural function evaluations, limiting low-latency speech synthesis. MeanFlow reduces the sampling budget by modeling the average velocity over a temporal interval, yet maintaining high acoustic quality under extremely few-step token-to-mel generation remains challenging. To address this challenge, we propose X-Pred MeanFlow, a few-step streaming token-to-mel decoder that reparameterizes MeanFlow with mel-space prediction. The decoder predicts a generalized mel field and analytically derives the corresponding average velocity for sampling, thereby preserving the MeanFlow formulation while providing a direct acoustic prediction target. We further introduce layer-selective block-wise attention to enable continuous chunk-wise generation with bounded context. Experiments show that X-Pred MeanFlow improves few-step token-to-mel synthesis over Direct-$u$ MeanFlow and supports stable streaming generation. Speech samples are available.https://renxiaming.github.io/xpred-meanflow-stream-demo
18. MP-Bench: Evaluating Voice Agents as a Multiparty Conversation Participant
Authors: Yi-Jen Shih, Shih-Yun Shan Kuan, Guan-Ting Lin, Kai-Wei Chang, Siddhant Arora et al.
Categories: eess.AS, cs.AI, cs.CL | Accepted to EMNLP 2026 Findings
Conversational voice agents have advanced significantly, offering increasingly natural human-machine interactions through both cascaded and end-to-end architectures. However, while recent benchmarks extensively evaluate dyadic interactions and passive audio comprehension, they largely overlook a prevalent real-world scenario: multi-party conversations. Evaluating agents in these settings is fundamentally more challenging than in dyadic interactions due to the exponentially greater conversational complexity. For voice agents to integrate seamlessly into human group dynamics, they must not only generate contextually appropriate responses but also demonstrate a nuanced understanding of open turn-taking. To address this gap, we introduce Multiparty Bench (MP-Bench), the first benchmark specifically designed to objectively evaluate conversational speech systems as active participants within multi-party contexts. MP-Bench assesses agent behavior along two primary dimensions: turn-taking awareness and response appropriateness. Additionally, we incorporate comprehension-based question-answering tasks as a complementary evaluation. By benchmarking 12 voice agents, we find that real-time voice agents stay at or below 22% on multiparty comprehension and remain near chance on multiparty turn-taking, exposing an open challenge for real-time voice agents under multiparty scenario.
19. Overview and Meta-Analysis of DCASE 2026 Challenge Task 6: Audio Moment Retrieval from Long Audio
Authors: Hokuto Munakata, Tatsuya Komatsu, Keisuke Imoto, Taichi Nishimura, Huang Xie et al.
Categories: eess.AS, cs.SD
This paper presents an overview of the Detection and Classification of Acoustic Scenes and Events (DCASE) 2026 Challenge Task 6, Audio Moment Retrieval (AMR) from Long Audio. Given a several-minute-long audio recording and a free-form text query, AMR aims to retrieve temporal moments in the recording that match the query, where each moment is represented by a pair of start and end timestamps. This task requires effective cross-modal alignment and long-range temporal modeling. We describe the task definition, the evaluation metrics, the development and evaluation datasets, and a baseline system that combines a pre-trained MS-CLAP feature extractor with a Detection Transformer (DETR)-based moment-detection network. On the development data, the baseline trained on a manually annotated dataset and a synthetic dataset achieved Recall1@0.7 of 13.56%, indicating that AMR in long audio remains a challenging problem. The challenge attracted 21 teams, which submitted 59 systems in total. The three best systems achieved Recall1@0.7 of 48.59%, roughly 3.5 times the baseline score. The results show that strengthening the audio-text feature extractor and the moment-detection network led to substantial performance improvements. Furthermore, the top three teams boosted performance by applying confidence score calibration or ensembling across different temporal resolutions of features.
20. StepAudio 3 Music Technical Report
Authors: Chengli Feng, Zhiyue Wu, Jiahao Song, Zheqi Dai, Boyang Wang et al.
Categories: eess.AS, cs.SD | 18 pages, 6 figures. Audio demonstrations: https://stepaudiollm.github.io/step-audio-3-music
We introduce StepAudio 3 Music, a large-scale, long-form music generation model that supports explicit musical planning and open-domain text-controlled generation. The StepAudio Music Tokenizer represents audio as a 50-Hz stream from a 65536-entry single codebook, using semantically informed self-supervised and multi-task training to preserve musical structure and reconstruction-relevant information. A flow-matching diffusion Transformer (DiT) predicts continuous StepAudio VAE latents, which our VAE decoder converts into 48-kHz audio. This discrete-continuous design is guided by comparisons of single-codebook VQ, Semantic and Acoustic RVQ, and different DiT configurations. For explicit planning, a Mixture-of-Experts autoregressive model uses ABC notation to produce an intermediate arrangement plan (ABC-CoT) before predicting music tokens, making harmony, rhythm, and melodic structure part of the generation context. A progressive training curriculum and supervised fine-tuning support song and instrumental generation, accompaniment generation from dry vocals, and cover-song synthesis for up to 5 minutes and 30 seconds. With reinforcement learning via direct preference optimization (DPO), the final model achieves the highest AudioBox Content Enjoyment, Content Usefulness, and Production Quality scores and the highest MuQ-MuLan similarity among the evaluated systems, with competitive SongBench results. On the preliminary Artificial Analysis Music Arena Vocals leaderboard, it obtains a Quality Elo of 1105, behind only Suno V5.5 and Mureka and ahead of Suno V5, MiniMax models, and other systems. Audio demonstrations are available at https://stepaudiollm.github.io/step-audio-3-music.
21. $τ$-Elicitation: Benchmarking multi-turn entity extraction in voice agents
Authors: Soham Ray, Victor Barres
Categories: cs.AI
Voice agents often need to collect names, addresses, identifiers, dates, and times exactly, yet end-to-end benchmarks obscure where capture fails. We introduce $τ$-Elicitation, a 200-task voice benchmark spanning 10 entity types, controlled difficulty, caller realisms, and three environments. A matched text agent passes all tasks, but four voice configurations achieve robust exact success from 0.14 to 0.41. Agents increase verification for hard and unfamiliar entities and sometimes for incorrect captures, but not for their weakest caller voice; only 24 to 37 percent of verified errors are repaired. A scaffold that prescribes spelling, read-back, correction, and confirmation raises robust Pass$^3$ by 14 to 31 points, at a cost of 21 to 28 seconds per call. Realisms such as spelling variations and restarts do not detectably affect exact success; mispronunciation increases repair effort. These results identify strategy selection and successful recovery as the central bottlenecks in exact spoken entity collection.
22. SteerDuplex: Steerable Duplex Speech Dialogue Models
Authors: Utkarsh Tyagi, Ramaneswaran Selvakumar, Advait Gosai, Sonal Kumar, Nikhil Barhate et al.
Categories: cs.AI, cs.CL | 24 pages, 7 figures
Full-duplex spoken dialogue models support low-latency turn taking, interruption handling, and backchanneling, yet a key capability remains underexplored: steerability, the ability to reliably shift conversational behavior along attributes such as tone, persona, speaking rate, and voice style in response to user instructions. We introduce a taxonomy of text- and audio-based steerability that identifies substantial gaps in current full-duplex models. To address this gap, we introduce SteerDuplex, a Moshi-based full-duplex speech model fine-tuned on natural conversations and synthetic dialogues targeting instruction following, vocal delivery, reasoning, and duplex interaction. We further apply two-stage reinforcement learning (RL) with hybrid rewards, combining verifiable interaction checks and judge-based semantic feedback to improve timing and response continuity. To evaluate full-duplex spoken steerability, we introduce SteerBench, a benchmark with 390 spoken prompts and 1,067 human-authored binary audio and text rubrics spanning tone, persona, style/accent, and speed/length. On SteerBench, supervised training improves audio-steering average pass rate by 44.5 percentage points over the strongest evaluated open baseline. On Audio MultiChallenge, task average pass rate improves by 7 points over its strongest evaluated open baseline. RL further raises source-clean interruption response from 72.5% to 82.5% and reduces synthetic pause barge-in from 26.5% to 9%. Steering and aggregate task scores remain comparable or higher, while reward probes reveal reward hacking through incomplete responses. Our model and benchmark support systematic research on spoken steerability, with reward analysis showing why timing gains must be evaluated alongside response completeness.
23. Causal Analysis and Mitigation of Spurious Onsets in Full-Duplex Speech LLMs
Authors: Kento Nishi
Categories: cs.CL
Speech-to-speech LLMs like Moshi, and its derivative PersonaPlex, can listen and speak concurrently through full-duplex generation. However, they can begin speaking inappropriately during prolonged user silence: under digital-zero input, Moshi and PersonaPlex initiate speech in 30% and 27.5% of five-minute continuations, respectively. What causes this spurious speech? We investigate two hypotheses: either repeated sampling selects speech despite persistently low onset probabilities, or self-conditioning on nonspeech outputs causes an abrupt spike in onset probability. We find that, at every observed onset, speech probability spikes by over nine orders of magnitude in one 80-ms frame, supporting the latter hypothesis. Then, to suppress these onsets without blocking genuine responses, we ask a causal counterfactual question: is the model responding to user speech, or would its next-token distribution remain similar if the preceding user input were muted? Accordingly, we suppress onsets whose distributions change little under this intervention. Under realistic microphone noise, our method suppresses spurious onsets, while preserving genuine responses: one-sided 95% lower confidence bounds are 98.68% and 98.82% for Moshi, and 96.90% and 99.25% for PersonaPlex. Our inference-time method runs in real-time without retraining, with 95th-percentile decision time below 61 ms, within the 80-ms frame budget. Our code is available at https://github.com/KentoNishi/icassp27-spurious-onsets.
24. DuplexDrama: A Synthesized Dialogue Dataset with Scenarios, Full-Duplex Behaviors, Expressive Speech, and Sound Events
Authors: Qingxiang Guo, Wenke Fan, Shuofeng Zhao, Dawei Yang, Zhiyang Zhou et al.
Categories: cs.CL | 5 pages, 5 figures, 5 tables, 18 references. Demo: https://dunjie5465.github.io/duplexdrama-demo/
We present DuplexDrama, the first synthesized spoken dialogue dataset that simultaneously covers four dimensions: (i) complete persona and scenario settings; (ii) three full-duplex behaviors (interruption, backchannel, incomplete); (iii) expressive speech with persona-aligned emotion labels; and (iv) script-aware sound events. DuplexDrama is built via a 4-stage pipeline; quality validation on both scripts and synthesized audio confirms its quality. We have produced more than 2,000 hours audio data with a 64-voice timbre pool spanning 13 personas and 5 age buckets; 3.8% of all turns carry at least one full-duplex behavior. This data has been validated through internal full-duplex model training. We will release a curated subset of 6,400 bilingual dialogues (800 h, Chinese ~500 h + English ~300 h) to advance full-duplex spoken dialogue model research. Data samples are available at our demo page and LLM-judge evaluation prompts will be released with the dataset.
25. Not All Speech Is Intent: Adaptive Self-Correcting Inference Layer for Post-ASR False Wake-Up
Authors: Preeti Saraswat, Divya Neelagiri, Anil Yadav
Categories: cs.CL, cs.AI
False wake-up activations remain a persistent challenge in conversational AI. Speech phonetically similar to a device’s wake word can produce a syntactically valid and semantically coherent ASR transcript that the assistant incorrectly executes. Most existing systems make a single intent decision in isolation, without a mechanism to learn from recurring errors over time or adapt to individual users through personalized learning. We introduce the Feedback-Driven Adaptive Self-Correcting Inference Layer (ASCIL), a complementary post-ASR correction framework that re-evaluates wake-up intent before response generation by fusing acoustic embeddings, linguistic cues, device context, and patterns from past misclassifications. ASCIL interprets implicit signals, including hesitation, disengagement, and silence, and explicit signals, including cancellation and repetition, as automatically inferred, noisy behavioral indicators of potential misclassification. These signals drive online pattern updates without manual annotation, whereas the intentional/unintentional reference labels used for offline evaluation are human-annotated. It generalizes from prior errors, applies corrective adjustments at inference time, and continuously updates in parallel with natural-language execution. Evaluated on a proprietary dataset of 3,667 interactions with human-annotated intentional/unintentional reference labels spanning 14 acoustic and contextual conditions, ASCIL achieves 54.27% relative error reduction on a session-disjoint subset constructed from baseline failures, and up to 24.39% relative error reduction at threshold 0.90 on the issue-tagged evaluation slice. These gains are achieved while improving intentional acceptance rates, with a median added latency below 60 ms in the reported benchmark.
26. Physical-State-Guided Diffusion Sampling for Full-Waveform Inversion
Authors: Chen Min, Haowen Jiang, Zheng Ma, Xiongbin Yan
Categories: cs.LG | 43 pages, 13 figures
Full waveform inversion (FWI) estimates subsurface velocity from seismic recordings, but its ill-posedness and nonlinearity make accurate reconstruction strongly dependent on initialization and prior information. Diffusion posterior sampling provides a learned geological prior, yet directly coupling its denoiser to the nonlinear wave solver can yield unreliable physical guidance. We propose Physical-State-Guided Diffusion Sampling (PSG), which couples a persistent physical velocity to the diffusion prior through a Gaussian bridge. The physical state is refined by waveform fitting regularized by the denoised velocity, and in turn guides the reverse diffusion process. This formulation separates the wave-equation and denoiser gradients while preserving conventional FWI initialization and accumulated optimization history. On four OpenFWI families, PSG’s terminal denoised estimates outperform classical and diffusion-based baselines under clean and missing-trace acquisitions and maintain strong structural recovery under measurement noise. Repeated stochastic runs preserve the dominant geological structures, with ensemble variability concentrated near geological interfaces and positively associated with local inversion error. A frozen OpenFWI-trained prior further supports inversion of the larger Marmousi, Overthrust, and BP2004 Salt models, recovering complex geological structures without retraining.
27. Audiovisual diarization of overlapping click trains in sperm whale (Physeter macrocephalus) vocal sparring using a three-hydrophone array
Authors: Lara Berkenbaum, Hervé Glotin, François Sarano, Walter M X Zimmer, Véronique Sarano et al.
Categories: q-bio.QM, eess.SP
The individual attribution of sperm whale (Physeter macrocephalus) vocalizations during surface interactions constitutes a methodological challenge due to acoustic overlaps, multipath propagation, body shadowing, and indistinguishable inter-pulse intervals among similar-sized individuals. A multimodal audio-visual workflow is presented to deinterleave and attribute uncharacterized click trains produced by immature males during ``vocal sparring’’ using a portable three-hydrophone array coupled with synchronized video. The approach combines spectro-temporal tracking, based on inter-click interval dynamics and the Constant-Q Transform, with spatial time-difference-of-arrival modeling projected onto the image plane for optical validation. Analysis of 2,655 manually validated clicks shows that purely acoustic clustering diarization errors remain low, peaking at 25.79% only during extreme temporal superpositions. Integrating the optical modality resolves residual spatial indeterminacies; although near-planar array geometry induces vertical ambiguities, the system achieves up to 100% horizontal visual concordance for primary emitters. Crucially, this framework successfully reconstructed and assigned 11 distinct, intertwined click trains totaling 882 clicks to specific focal individuals despite near-field tactile constraints. Because ethological descriptions remain incomplete without identifying the emitter, explicitly correlating these emissions with physical kinematics provides the fine-scale resolution required to define this socio-acoustic behavior.