arXiv_SD
-
Automatic Contextual Audio Denoising
2026-05-21 10:06:34Diep Luong, Konstantinos Drossos, Mikko Heikkinen, Tuomas VirtanenAbstract
Audio context determines which sound components and sources are relevant and which can be perceived as irrelevant (noise) by listeners. For example, traffic noise is informative in urban surveillance but noise for a phone call at the same location. Most current audio denoising systems apply fixed target-noise definitions, often removing useful components in one context while failing to suppress irrelevant components. To address this, we introduce the concept automatic contextual audio denoising (ACAD) which defines target and noise based on the inferred context. In this work, we restrict context to be associated with an acoustic scene class. We label sound events outside the event distribution of a scene class (noise) as out-of-context (OC) and events typical for that scene as in-context (IC). We implement a deep learning method that automatically infers the context of the audio signal and removes OC components, and benchmark it against variants: without context inference, with oracle context, and with separately provided uninformative context. On paired clean/noisy data across diverse contexts, where OC components in one context may be IC in another, our proposed method outperforms other approaches across standard objective metrics, indicating that the model can infer context and context-dependent processing can enhance denoising.
Abstract (translated)
URL
https://arxiv.org/abs/2605.22262
PDF
-
Effective User-defined Keyword Spotting with Dual-stage Matching, Multi-modal Enrollment, and Continual Adaptation
2026-05-21 07:52:21Zhiqi Ai, Han Cheng, Shiyi Mu, Xinnuo Li, Yongjin Zhou, Shugong XuAbstract
User-defined keyword spotting (KWS) is crucial for personalized voice interaction, yet existing methods face several challenges: (1) insufficient discriminability among confusable words, (2) performance inconsistency across speakers with varying pronunciations, and (3) high data cost to ensure reliable wake-word performance. In this paper, we introduce DMA-KWS, an efficient and robust framework for user-defined keyword spotting. First, it adopts a dual-stage matching pipeline: CTC decoding with streaming phoneme search to locate candidate segments, followed by QbyT with a phoneme matcher for fine-grained verification, enabling it to better distinguish confusable words. Next, multi-modal enrollment fuses user-specific speech with text embeddings to further improve accuracy for registered users. Finally, a parameter-efficient continual adaptation mechanism performs lightweight updates using synthetic and real data. Extensive experiments demonstrate the superior performance of DMA-KWS. On the LibriPhrase Hard subset, it achieves 97.85% AUC and 6.13% EER, reaching state-of-the-art performance. In speaker-dependent settings, DMA-KWS consistently outperforms text-only enrollment, demonstrating significant performance gains. Moreover, the proposed parameter-efficient fine-tuning mechanism adapts DMA-KWS with only 187k updated parameters, further enhancing KWS performance while ensuring suitability for on-device deployment.
Abstract (translated)
URL
https://arxiv.org/abs/2605.22120
PDF
-
RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching
2026-05-21 07:22:28Jinhyeok Yang, Hyeongju Kim, Yechan Yu, Joon Byun, Frederik Bous, Juheon LeeAbstract
While flow-matching text-to-speech (TTS) achieves strong zero-shot speaker similarity and naturalness, it remains susceptible to content fidelity issues, particularly skip and repeat errors from imperfect alignment. We propose RobustSpeechFlow, a training strategy that improves alignment robustness by extending contrastive flow matching with length-preserving repeat and skip latent augmentations. Requiring no external aligners or preference data, our method directly penalizes realistic failure modes and readily integrates into existing pipelines. On Seed-TTS-eval, it reduces the word error rate (WER) from 1.44 to 1.38 using only 0.06B parameters. On our ZERO500 benchmark, it delivers consistent intelligibility improvements across diverse speaker and prosody conditions; at NFE=24, it reduces English character error rate (CER) from 0.48\% to 0.35\% and Korean CER from 0.81\% to 0.57\%. Audio samples: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2605.22083
PDF
-
Real-time, EDM-inspired sonfication of the activity of a supercomputer
Abstract
The project described in this paper explores the informative sonification of data received in real time from a supercomputer. These data capture the current activities in all the nodes of the computer, therefore, their sonification functions as a form of continuous monitoring of the nodes' behavior and, by extension, of the system as a whole. Because such monitoring is theoretically unending, the resulting sonification must be musically capable of conveying information through sound in a way that remains both intelligible and engaging over long durations. Rather than imposing a predefined musical style onto the data, we sought to identify one which the data themselves could plausibly support. From a small set of candidates, we selected EDM because it is a family of genres whose structural and temporal characteristics align well with continuous, data-driven processes and long-term listening. Through this style-based approach, this research builds on the long tradition of computer data sonification while uniquely combining three elements rarely addressed together: monitoring (rather than debugging) as the primary goal, real-time (rather than post-mortem) data interpretation, and generation of virtually infinite and stylistically coherent (rather than incongruous) music.
Abstract (translated)
URL
https://arxiv.org/abs/2605.21874
PDF
-
Instrumental Text-to-Music Generation with Auxiliary Conditioning Branches
Abstract
Text-to-music generation has advanced rapidly, with modern autoregressive and diffusion-based models producing convincing music from natural-language prompts. However, much of this progress relies on large-scale training data and external pretraining, making it difficult to isolate which design choices remain effective when data and pretraining are controlled. We study this setting using a Diffusion Transformer backbone with lyric and timbre conditioning, adapted to an instrumental-only text-to-music task in which the auxiliary lyric and timbre branches receive only degenerate conditioning signals. Through controlled ablations, we find that models retrained without these branches score lower across AudioBox aesthetics, LLM-as-judge, and human MOS, and that reinvesting the saved parameters as additional DiT depth recovers only marginally. This suggests the auxiliary branches may act as training-time architectural anchors whose contribution goes beyond their explicit conditioning content. We validate the same model through comparisons with external instrumental baselines and through our submission to the ICME 2026 Academic Text-to-Music (ATTM) Grand Challenge, where our Performance submission ranked first under both the objective metrics and the subsequent organizer-administered MOS over 35 raters, attaining the highest overall MOS across all challenge submissions, while our Efficiency submission was a finalist that tied for second under the objective metrics.
Abstract (translated)
URL
https://arxiv.org/abs/2605.21433
PDF
-
CoarseSoundNet: Building a reliable model for ecological soundscape analysis
2026-05-20 13:18:11Alexander Gebhard, Andreas Triantafyllopoulos, Dominik Arend, Sandra M\"uller, Svenja Schmidt, Michael Scherer-Lorenzen, Bj\"orn W. SchullerAbstract
A soundscape is composed of three types of sound: biophony (sounds made by animals), geophony (natural abiotic sounds) and anthropophony (sounds made by humans). A key research question in the field of soundscape ecology is how these components interact with each other, specifically how biophony responds to geophony and anthropophony. Nevertheless, as of today, there are not many analytical instruments that enable the distinct quantification of these elements. Recent machine learning (ML) approaches aim to support automated analysis but often rely on task-specific or clean data, limiting generalisation to noisy passive acoustic monitoring (PAM) recordings. This study presents a clear and reproducible structure to build ML models for coarse soundscape classification and introduces CoarseSoundNet, a deep learning model trained to distinguish biophony, geophony, and anthropophony under realistic PAM conditions. We systematically investigate model architectures, the influence of an additional training class, data composition, and evaluation strategies. Our findings suggest that model performance improves with additional PAM data, especially when similar to the target domain, and by introducing an explicit silence class during training. Class-specific decision thresholds and duration-based constraints further enhance performance, particularly for anthropophony and geophony. Error analyses exhibit challenges for anthropophony due to masking effects and confusions for silence and insect sounds for geophony and biophony. Finally, we conduct an ecological case study which shows that pre-filtering recordings with CoarseSoundNet yields acoustic index trends comparable to ground-truth filtering, supporting its use as an effective preprocessing tool for ecoacoustic analyses.
Abstract (translated)
URL
https://arxiv.org/abs/2605.21143
PDF
-
Musical Attention Transformer: Music Generation Using a Music-Specific Attention Model
2026-05-20 12:16:28Shinnosuke Taksuka, Hideo MukaiAbstract
This study aims to enhance the quality of music generation using Transformers by incorporating meta-information. While Transformer-based approaches are effective at capturing long-term dependencies in musical compositions, the music they generate often suffers from issues such as excessive repetition or duplication of notes, leading to unnatural melodies. To address these limitations, we propose Musical Attention, a mechanism that incorporates meta-information such as bar numbers, key, signatures, and tempos into the attention process. Musical Attention explicitly leverages both the structural properties of music and its associated metadata, enabling the Transformer's attention mechanism to operate more effectively and thereby improving the quality of the generated output. In our framework, each musical note is represented as a combination of five events-pitch, bar number, onset, duration, and velocity in addition to the three metadata elements. The attention mechanism is then modified to reflect the correlations among these eight features, allowing the model to better capture the inherent characteristics of musical composition. Experimental results demonstrate that the model incorporating Musical Attention outperforms prior methods, such as Full Attention and Strided Attention, in terms of musical coherence, variation, and overall quality. Notably, it significantly reduces repetition and enhances the model's ability to generate diverse, harmonically consistent melodies. Musical Attention thus represents a meaningful advancement in AI-driven music generation, facilitating the creation of more natural and expressive compositions.
Abstract (translated)
URL
https://arxiv.org/abs/2605.21081
PDF
-
DUET: Unified Dual-Space Emotion Control for Diffusion and Flow-Matching Driven Text-to-Speech
2026-05-20 07:46:16Xu Zhang, Longbing Cao, Zhangkai WuAbstract
Diffusion and flow-matching based text-to-speech (TTS) models excel in naturalness but often lack explicit emotion control, as emotional signals remain entangled with speaker identity. We discover that emotion embedding emerges as a linearly decodable direction of frozen hidden states, nearly orthogonal to the direction embedding speaker identity. This inspires a plug-and-play framework DUET for emotion control over pretrained diffusion and flow-matching based TTS models. During generation, DUET unifies dual-space control to achieve fine-grained emotion intervention in a single per-step update: hidden space steering shifts generation along the target emotion direction, while mel-space guidance refines spectral details through gradients backpropagated from a differentiable vocoder. We validate DUET on five architecturally diverse pretrained TTS backbones across three datasets, where it outperforms 10 supervised state-of-the-art emotional TTS baselines across paradigms and achieves the highest human-rated emotion appropriateness. To further showcase its qualitative behavior, we deploy DUET on an Ameca humanoid robot, where it produces richly expressive emotional speech on the humanoid, demonstrating the strong potential for plug-and-play affective interaction for embodied agents.
Abstract (translated)
URL
https://arxiv.org/abs/2606.00066
PDF
-
SEABAD: A Tropical Bird Activity Detection Dataset for Passive Acoustic Monitoring
2026-05-20 07:44:39Muhammad Mun'im Ahmad Zabidi, Mohd Yamani Idna Idris, Norisma IdrisAbstract
Passive acoustic monitoring (PAM) enables large-scale biodiversity assessment, but continuous recording generates large amounts of non-informative audio, creating challenges for storage, power consumption, and long-term edge deployment. Bird audio detection (BAD), which identifies bird vocalizations, can reduce this burden by filtering irrelevant recordings before downstream analysis. However, most BAD systems are trained on temperate datasets despite tropical soundscapes being denser, more species-rich, and acoustically unpredictable. To address this gap, we introduce SEABAD (Southeast Asian Bird Activity Detection), a dataset of 50,000 curated three-second clips from Southeast Asian soundscapes, evenly balanced between bird-present and bird-absent samples. The dataset spans 1,677 bird species and is standardized to 16 kHz mono audio for embedded and low-power inference. We developed a dual-branch curation pipeline: a six-stage positive-label workflow applied to Xeno-Canto recordings, alongside six source-specific negative-label extractions from environmental datasets. These procedures reduced class imbalance by 13.7% (Gini coefficient: 0.601 to 0.519). A manual audit of 1,000 positive clips confirmed 97.8% +/- 0.9% labeling accuracy. Baseline experiments using MobileNetV3-Small achieved 99.57% +/- 0.25% accuracy and 0.9985 +/- 0.0002 AUC across three random seeds. SEABAD and the full curation pipeline are publicly released to support tropical BAD research and energy-efficient acoustic monitoring.
Abstract (translated)
URL
https://arxiv.org/abs/2605.20853
PDF
-
Academic Text-to-Music Grand Challenge: Datasets, Baselines, and Evaluation Methods
2026-05-20 07:18:24Fang-Chih Hsieh, Wei-Jaw Lee, Chun-Ping Wang, Hung-yi Lee, Hao-Wen Dong, Yi-Hsuan YangAbstract
This paper presents an overview and the technical framework of the ICME 2026 Grand Challenge on Academic Text-to-Music Generation (ATTM). Despite the rapid progress in text-to-music generation (TTM) systems, the field is currently dominated by models trained on massive proprietary datasets with industrial-scale computational resources, creating a significant barrier for academic research. To address this, the ATTM Challenge establishes a fair-play benchmark that requires participants to train generative models strictly from scratch using a standardized, CC-licensed subset of the MTG-Jamendo dataset containing only instrumental music. The challenge is divided into two tracks: the Efficiency Track (limited to 500M parameters) and the Performance Track (no parameter limit). Submissions are evaluated through a multi-stage process involving objective metrics, including Frechet Audio Distance, CLAP score, and a novel Concept Coverage Score (CCS), followed by a subjective listening test. By providing open-source baselines, preprocessing pipelines, reference captions, and public evaluation code for computing FAD and CLAP, this challenge aims to facilitate and promote TTM research in academic contexts.
Abstract (translated)
URL
https://arxiv.org/abs/2605.21538
PDF
-
A strongly annotated passive acoustic dataset for tropical bird monitoring
2026-05-20 00:36:06Daniela Ruiz, Juan Sebasti\'an Ulloa, Zhongqi Miao, Nicol\'as Betancourt, Maria Paula Toro-G\'omez, Andr\'es Hern\'andez, Bruno Demuro, Eliana Barona-Cort\'es, Angela Mendoza-Henao, Andr\'es Sierra-Ricaurte, Sebasti\'an P\'erez-Pe\~na, Rahul Dodhia, Pablo Arbel\'aez, Juan M. Lavista FerresAbstract
Passive acoustic monitoring enables continuous, non-invasive biodiversity assessment across diverse ecosystems. The scale of these datasets has driven the adoption of machine learning, with supervised approaches showing strong performance. However, supervised methods require time-resolved annotated datasets, which remain scarce, especially in complex tropical soundscapes. We present PteroSet, a curated dataset of strongly annotated Neotropical bird vocalizations recorded in Puerto Asis (Putumayo) and Pivijay (Magdalena), Colombia, between 2023 and 2025. The dataset comprises 563 recordings (73.62 h) and 15,372 time-frequency annotations, including 6,702 events identified to the species level across 168 species. We release the annotations in a COCO-inspired JSON schema that unifies audio files, taxonomic categories, and labels for machine learning workflows. Beyond providing annotated data, PteroSet serves as a realistic benchmark that highlights key characteristics of tropical soundscapes, including acoustic co-occurrence and domain shift across recording sites. We provide a deep learning baseline for binary bird detection, demonstrating PteroSet's usability and the challenges it presents.
Abstract (translated)
URL
https://arxiv.org/abs/2605.20578
PDF
-
Codec-Robust Attacks on Audio LLMs
2026-05-19 21:39:52Jaechul Roh, Jean-Philippe Monteuuis, Jonathan Petit, Amir HoumansdarAbstract
Prior attacks on Audio Large Language Models (Audio LLMs) demonstrated that carefully crafted waveform-domain perturbations can force targeted adversarial outputs. As a defense mechanism against these attacks, real-world codec compression preprocessing has been studied to both detect and remove the perturbations. Yet no existing attack has demonstrated robustness against these compressions. We introduce CodecAttack, which optimizes a perturbation in a neural audio codec's continuous latent space rather than directly perturbing the audio waveform. We show that the codec's compression channel, which discards waveform perturbations, transmits perturbations crafted in its own latent space. To further harden the attack across real-world compression channels, we apply multi-bitrate straight-through Expectation-over-Transformation (EoT), all without modifying the target model. Across three realistic Audio LLM deployment scenarios and three target models, CodecAttack achieves an average 85.5% target-substring attack success rate (ASR) on Opus at moderate bitrates, while the waveform baseline trained with identical EoT hardening does not exceed 26% at any bitrate. The attack transfers to held-out codecs, reaching up to 100% ASR on MP3 and 84% on AAC-LC without retraining. A per-band energy analysis shows that the latent perturbation concentrates below 4kHz, exactly where codecs allocate the most bits, while the waveform baseline spreads into higher frequencies that codecs discard. These results demonstrate that lossy compression is not a reliable defense against adversarial audio and that codec-aware attacks pose a practical threat to deployed Audio LLM systems.
Abstract (translated)
URL
https://arxiv.org/abs/2605.20519
PDF
-
Music of Changing Lines: Toward a Culturally Situated Approach to the I-Ching
2026-05-19 18:35:47Ling Qi, Aleksandra Teng Ma, Alexandria SmithAbstract
The I-Ching is one of the most influential texts in Chinese intellectual history, integrating divination, cosmology, and ethical reflection. While Western experimental music, most notably John Cage, has drawn on the I-Ching as a source of chance operation, such appropriations have often detached its formal mechanisms from the interpretive and philosophical processes that give the text meaning. This work, Music of Changing Lines, presents an interactive system that re-centers the I-Ching as a meaning-bearing framework rather than a neutral randomizer. Users perform Wen Wang Fa coin casting, which is accompanied in real time through probabilistic musical processes. The resulting hexagrams and changing lines are interpreted by a large language model, Gemini, in relation to the user's inquiry. This textual interpretation is then translated into a prompt for a generative music model, Lyria, producing a responsive musical realization. By situating AI as an interpretive intermediary rather than a compositional authority, the system foregrounds the I-Ching's ritual, interpretation, and participation as the primary sonic materials. Music of Changing Lines extends process-driven traditions in computer music by demonstrating how generative AI can support participatory, meaning-driven musical processes without prescribing musical structure or replacing human agency.
Abstract (translated)
URL
https://arxiv.org/abs/2605.20386
PDF
-
Precise and Simple Audio-to-Score Alignment
Abstract
Audio-to-score alignment is a long-standing challenge in music information retrieval and arguably the most widely applicable alignment task for music research. Alignment algorithms match two versions of a piece of music, and for this to work these versions need to be in comparable formats. Audio-to-audio alignment matches audio features; when matching audio files to scores, they must either synthesize the score or derive audio-like features by means of piano rolls or similar feature sequences. Symbolic alignment, by contrast, matches symbolically encoded notes; in an audio-to-score scenario these would be obtained by a transcription of the audio file. In this article, we present an algorithm that bridges audio-like and symbol-level features directly. Sequential audio features encoding onset and spectral activation are matched to score positions by a bespoke dynamic programming-based matching algorithm derived from symbolic alignment methods. The resulting method is both precise - surpassing widely used audio-to-audio approaches based on synthesized scores -, and remains flexible in its digital signal processing components, i.e., the method is adaptable to diverse timbral characteristics without requiring a separate transcription model. Furthermore it inherits some of the symbolic alignment runtime advantages with an algorithmic complexity that is at worst linear in the length of the (typically short) symbolic score and (typically long) audio feature sequence. In the following sections, we provide a detailed algorithm description and evaluate its alignment quality on a large-scale dataset of solo piano recordings.
Abstract (translated)
URL
https://arxiv.org/abs/2605.20014
PDF
-
A conceptual framework for learning to listen by reward: Curiosity-driven search for novel sources
2026-05-19 15:24:18Andreas Triantafyllopoulos, Jakub \v{S}\v{t}astn\'y, Alexios Terpinas, Tianyi Liu, Yuanqi Wang, Bj\"orn W. SchullerAbstract
Reinforcement learning is a powerful learning paradigm that has spearheaded progress in numerous domains. Its core promise lies in learning through high-level goals without the need for granular labels. However, it still remains elusive in the realm of audio, where it has received substantially less attention than in computer vision or other domains. The key question remains: how can agents learn to listen purely via reward-driven exploration? In this contribution, we present an overview of previous attempts and a new conceptual framework for learning to listen by reward. Our approach depends on the continuous search for novel sound sources. We formulate our framework, discuss open technical challenges, and present a first proof-of-concept implementation that showcases the feasibility of our approach.
Abstract (translated)
URL
https://arxiv.org/abs/2605.19984
PDF
-
DASM: Domain-Aware Sharpness Minimization for Multi-Domain Voice Stream Steganalysis
2026-05-19 15:08:41Pengcheng Zhou, Pianran Guo, Shuhua Chen, Mengqin Zhao, Zhongliang Yang, Linna ZhouAbstract
The growing use of information hiding in network streaming media for covert communication poses a significant security threat, necessitating the development of robust detection technologies. However, existing steganalysis methods for network voice streams mostly rely on data distributions in specific scenarios, making it difficult to adapt to the practical detection needs of non-homologous data distributions. Through Hessian analysis, we find that the loss landscapes of mainstream models are dominated by numerous saddle points and sharp local minima, rendering them highly sensitive to data distribution shifts and fundamentally limiting generalization. Therefore, we propose a new optimizer, Domain-Aware Sharpness Minimization (DASM). The core mechanisms of DASM consist of two aspects: first, it integrates domain-supervised contrastive learning with sharpness-aware optimization, explicitly preserving inter-domain feature separation while seeking flat minima; second, we design an adaptive domain gap modulation strategy that dynamically calibrates the optimization loss weights by sensing the real-time feature separability of different domains. Extensive experimental results demonstrate that our method outperforms the state-of-the-art methods by a large margin and achieves excellent generalization and robustness.
Abstract (translated)
URL
https://arxiv.org/abs/2605.19955
PDF
-
Cross-Talk Speech Reduction, by Separation, for Separation
2026-05-19 11:29:35Zhong-Qiu Wang, Samuele CornellAbstract
In conversational speech separation and recognition tasks, close-talk microphones are typically attached to each speaker during training data collection to capture near-field, close-talk mixture signals, in addition to using far-field microphones to record far-field mixture signals. Each such close-talk mixture exhibits a reasonably high energy level for the wearer and could intuitively serve as weak supervision for training far-field speech separation models directly on real-recorded far-field signals. However, they are not sufficiently clean for this purpose, as they often contain strong cross-talk speech from other speakers in addition to background noise. To address this, we propose cross-talk reduction (CTR), a task aiming to isolate the wearer's speech from each close-talk mixture, and a novel method called CTRnet, which can be trained directly on real-recorded pairs of close-talk and far-field mixtures to accomplish CTR. Building on CTRnet, we further propose pseudo-label based far-field speech separation (PuLSS), which uses CTRnet's estimated clean speech as pseudo-labels to train models for separating far-field mixtures. A key advantage of the proposed framework is that both CTRnet and PuLSS can be trained on real-recorded data from the target domain, addressing the generalization gap commonly observed when models are trained exclusively on simulated data. On the CHiME-6 dataset, our framework achieves state-of-the-art ASR performance under both oracle and estimated speaker diarization, surpassing all CHiME-{7,8} challenge submissions. To our knowledge, it is the first neural speech separation method that substantially outperforms guided source separation on real conversational "speech-in-the-wild" data.
Abstract (translated)
URL
https://arxiv.org/abs/2605.19695
PDF
-
Executable Boundary Contracts for Sound Event Traces
Abstract
Sound event reports often compress timed boundary behavior into frame, segment, or event scores. This paper defines executable boundary contracts for finite sound event traces. The frame fragment is a bounded Boolean fragment embeddable in STL after grid projection. The event layer adds declared interval matching, duration clauses, fragmentation clauses, and obligation restricted vector scoring. The aim is measurement, not a new general temporal logic and not a challenge leaderboard. The artifact evaluates controlled Mini LibriSpeech seeded scenes, MAESTRO Real soundscapes, frozen pretrained timing probes, and an official DCASE 2024 Task 4 baseline track. Across these tracks, standard scores and contract coordinates disagree in interpretable ways. The strongest real corpus finding is that union activity can hide typed boundary failure, while external DCASE outputs provide a class indexed challenge level reference. Code, generated tables, manifests, and Lean checks for the finite frame core are supplied as ancillary material.
Abstract (translated)
URL
https://arxiv.org/abs/2605.19632
PDF
-
Optimising Neural Speech Codecs for 300bps Communication using Reinforcement Learning
2026-05-19 08:40:12Junyi Wang, Chi Zhang, Jing Qian, Haifeng Luo, Hao Wang, Zengrui Jin, Chao ZhangAbstract
In bandwidth-constrained communication such as satellite and underwater channels, speech must often be transmitted at ultra-low bitrates where intelligibility is the primary objective. At such extreme compression levels, codecs trained with acoustic reconstruction losses tend to allocate bits to perceptual detail, leading to substantial degradation in word error rate (WER). This paper proposes ClariCodec, a neural speech codec operating at 300 bit per second (bps) that reformulates quantisation as a stochastic policy, enabling reinforcement learning (RL)-based optimisation of intelligibility. Specifically, the encoder is fine-tuned using WER-driven rewards while the acoustic reconstruction pipeline remains frozen. Even without RL, ClariCodec achieves 4.64% WER on the LibriSpeech test-clean set at 300 bps, already competitive with codecs operating at higher bitrates. Further RL fine-tuning reduces WER to 3.55% on test-clean and 10.4% on test-other, corresponding to a 23% relative reduction while preserving perceptual quality.
Abstract (translated)
URL
https://arxiv.org/abs/2605.19541
PDF
-
Heterogeneity-Aware Dataset Scheduling for Efficient Audio Large Language Model Training
2026-05-18 20:41:08Yanru Wu, Jianning Wang, Chongxin Gan, Yang LiAbstract
Training general-purpose Audio Large Language Models (ALLMs) across diverse datasets is essential for holistic audio understanding, yet it faces significant challenges due to dataset heterogeneity, which often leads to conflicting gradients and slow convergence. Despite its impact, how to explicitly manage this heterogeneity during training remains underexplored, with current practices relying primarily on uniform mixture. In this work, we analyze multi-dataset AudioQA training from a convergence perspective and propose Grouped Sequential Training (GST). GST strategically organizes datasets into affinity-aware groups and introduces them via a progressive scheduling protocol, effectively balancing the stability of parallel training with the efficiency of sequential optimization. To ensure scalability, we develop gradient-based affinity metrics that capture inter-dataset relationships without the prohibitive cost of empirical transferability estimation. Extensive evaluations on 14 AudioQA datasets spanning speech, music, and environmental sounds demonstrate that GST achieves 30--40\% faster convergence than standard parallel training while maintaining or even surpassing the performance of mix-all training. Our results provide both theoretical insights and a practical, model-agnostic framework for efficient large-scale ALLM optimization.
Abstract (translated)
URL
https://arxiv.org/abs/2605.19101
PDF
