Self-supervised learning (SSL) yields powerful, context-rich representations for speech emotion recognition (SER), yet aggregating these representations into holistic descriptors remains a bottleneck. Conventional first-order aggregation implicitly assumes feature independence, which overlooks the latent Riemannian geometry and discards higher-order relationships essential to the representational power of the backbone. To address this problem, this paper proposes a novel Second-Order Correlation (SOC) layer. Instead of treating features in isolation, SOC models feature correlations as covariance descriptors to capture synergistic co-occurrence patterns, which serve as discriminative signatures for robust emotion recognition. By mapping these descriptors from the Riemannian manifold to a Euclidean tangent space through Log-Euclidean mapping (LEM), the proposed method preserves geometric integrity while enabling direct linear discriminative learning. Extensive experiments on the ESD and RAVDESS datasets demonstrate that SOC recovers discriminative information lost in first-order pooling and effectively aggregates high-dimensional SSL features.

Full-duplex spoken dialogue models allow voice agents to listen and speak concurrently, enabling natural interaction with real-time overlap. However, end-to-end dual-channel models that jointly encode user and agent streams may degrade in realistic acoustic environments: interfering speakers leaking into the user microphone can be encoded as part of the user query, corrupting the LLM's conditioning and causing unstable turn-taking and reduced response quality. We propose Interference-Resilient Adaptive Fusion (IRAF), a lightweight, streaming-compatible module that modulates the contribution of user audio to the LLM frame by frame. IRAF predicts a scalar reliability gate from target-speaker and user audio embeddings and rescales user representations before fusion with agent embeddings. Experiments on MS-MARCO and InstructS2S-200K show consistent gains in response quality and full-duplex interaction under interfering-speaker conditions.

Retrieving music using natural language descriptions has improved with contrastive audio-text models such as CLAP, but current systems remain limited to coarse semantic queries. When descriptions specify fine-grained musical attributes such as tempo, key, chord progression, or rhythmic structure, existing models often fail to retrieve the correct audio. We show that this limitation stems from the contrastive learning objective itself: despite being trained on long captions, CLAP-based models effectively utilize only the first few tokens, discarding much of the information encoded in detailed prompts. Then, we propose FIGMA (FIne-Grained Music RetrievAl), a multi-view contrastive architecture that addresses this limitation by jointly optimizing global audio-text alignment and frame-level, token-wise alignment. This design enables FIGMA to capture both high-level semantic context and fine-grained musical attributes within a unified representation space. Moreover, we formalize the task of Fine-Grained Music Retrieval and construct Fine-Grained Music Caption dataset (FGMCaps), a large-scale dataset of 380K music-caption pairs for training along with a 10K test set, both annotated with tempo, key, chord progression, beat count, as well as genre and mood. Extensive experiments demonstrate that FIGMA consistently outperforms existing CLAP-based music retrieval models across multiple music retrieval benchmarks, including out-of-domain evaluations, with relative improvements of up to 73.3%.

Discrete speech units obtained via k-means clustering of self supervised embeddings entangle phonetic, speaker, and language information, causing speaker mixing and cross-lingual interference in multilingual multi-speaker speech generation. Despite growing use in Audio LLMs and speech to speech systems, unit vocoders remain underexplored. We analyze a BigVGAN based unit vocoder, across four Indian languages. We study the interaction between cluster size and conditioning strategies using WER, speaker similarity, and unit level metrics. Results show that cluster size governs intelligibility by improving phonetic discriminability, while explicit speaker conditioning is indispensable for preventing identity collapse. Language supervision yields further gains mainly at lower cluster sizes where units remain ambiguous. Our analysis shows similar phonemes across languages collapse to the same cluster IDs at smaller inventories, with larger clusters progressively separating them.

The popularity of neural audio codecs as speech tokenizers has surged with the advent of Multimodal Large Language Models. New codec architectures with semantic and acoustic disentanglement have emerged. There are two main approaches to introduce semantic information into codec models: one distills semantic information from SSL representations into the first RVQ layer, while the other maintains separate streams for semantic and acoustic features. We propose HybridCodec, a unified architecture that combines both paradigms. It employs separate semantic and acoustic branches while distilling SSL representations into the semantic stream. This design ensures strong disentanglement without requiring an SSL model during inference. HybridCodec shows superior semantic specialization (RVQ-1) on in-domain test set and competitive reconstruction (RVQ-all). We demonstrate its robustness in out-of-domain and zero-shot cross-lingual settings, achieving a 3x speedup over existing dual-stream models.

Discrete speech tokens obtained from self-supervised learning (SSL) models provide efficient data compression while maintaining strong performance, and have been widely used as intermediate representations in various tasks. However, discretization inevitably causes information loss, leading to degraded performance compared with continuous SSL features. In this work, we propose to apply soft token assignment only during downstream inference. This approach preserves the efficiency of hard discretization during training while enhancing the expressiveness of the tokens at inference. The proposed method outperforms conventional hard assignment on both ASR and speech synthesis tasks, and exhibits particularly strong generalizability to out-of-domain data. For ASR of non-native speech, it even surpasses models using continuous SSL features. Moreover, analysis of the resulting representations shows they align more accurately with phonemes compared with conventional hard assignment.

We present VoxCPM2, a https://info.arxiv.org/help/prep#abstractsfully open-source multilingual and controllable speech generation foundation model that extends the hierarchical diffusion-autoregressive modeling paradigm of VoxCPM. VoxCPM2 advances the framework in three key dimensions: (i) capability, by unifying 30 languages, 9 Chinese dialects, natural-language voice design, style-controllable voice cloning, and high-fidelity continuation cloning within a single backbone; (ii) quality, through an asymmetric AudioVAE that encodes at 16 kHz and reconstructs at 48 kHz, enabling implicit super-resolution with high encoding efficiency; and (iii) scale, by jointly scaling the model to 2B parameters and the training data to over 2 million hours of multilingual speech. To support these diverse capabilities within one model, we introduce a unified sequence organization that expresses all generation modes through different arrangements of the same input building blocks, allowing joint training under a single set of parameters and objective. VoxCPM2 achieves state-of-the-art or competitive performance on public zero-shot and instruction-following TTS benchmarks. On our internal 30-language evaluation set, it attains an average WER of 1.68%. These results demonstrate that hierarchical continuous-latent modeling, without relying on any external discrete speech tokenizer, offers a viable and powerful foundation for large-scale multilingual and controllable speech generation. The model weights, fine-tuning code, and inference tools are publicly released under the Apache 2.0 license to foster community research and development.

Bioacoustic datasets from tropical regions remain limited, in part due to the absence of reproducible workflows for aggregating recordings from public archives. We present \textbf{MyGardenBird}, a curated dataset of bird vocalisations representing twelve common species across Peninsular Malaysia and the Indo-Malayan region. Recordings were sourced from Xeno-canto and processed through species-level filtering, manual spectrogram segmentation, and quality control checks. The primary release comprises 7,200 manually validated audio clips (16 kHz, 16-bit PCM mono WAV), balanced at 600 three-second clips per species (6.0 hours total) derived from 1,381 distinct recordings. Metadata includes geospatial coordinates, vocalisation categories, and signal-to-noise ratio (SNR) values (range: 0.83--59.18 dB; mean: 15.80 dB). A supplementary 44.1 kHz version is also provided. To mitigate data leakage, dataset partitions are defined at the source-recording level. Baseline classification experiments using convolutional neural networks on Mel-spectrograms achieved test accuracies of 92--96\%, indicating strong interspecies separability. Limitations include reliance on single-annotator curation; however, validation with BirdNET confirmed label consistency. MyGardenBird is openly available at https://doi.org/10.5281/zenodo.20306877 under a CC BY-NC-SA 4.0 licence. Complete preprocessing code accompanies the release to support reproducibility and future expansion.

We analyse error patterns of raw waveform acoustic models on TIMIT phone recognition beyond the overall phone error rate (PER). PER is decomposed across three broad phonetic class (BPC) categorisations, and confusion matrices are constructed from substitution errors. Our models combine parametric (SincNet, Sinc2Net) or non-parametric CNNs with Bidirectional LSTMs, achieving 13.9%/15.3% PER on Dev/Test, the best reported results for raw waveform models on TIMIT. Transfer learning from WSJ reduces PER to 11.3%/12.3%, surpassing the Filterbank baseline. Per-BPC analysis reveals that BLSTM layers benefit transition-dependent classes most, while WSJ transfer learning improves consonants roughly three times more than vowels. Confusion patterns are consistent across raw waveform and Filterbank systems, indicating that the dominant confusions reflect inherent phonetic similarities.

We present dots.tts, a 2B-parameter continuous autoregressive text-to-speech (TTS) foundation model that models speech in a continuous latent space. Compared with existing continuous autoregressive models, our key innovations are threefold. First, we train an AudioVAE with multiple objectives to build a semantically structured and prediction-friendly continuous speech space. Second, we use full-history conditioning in the flow-matching head to preserve long-range consistency and reduce drift during generation. Third, we apply reward-free self-corrective post-training to the flow-matching head to further improve robustness and acoustic quality. After being trained on a large-scale multilingual corpus, dots.tts achieves the best average performance on Seed-TTS-Eval, with WERs of 0.94%/1.30%/6.60% and SIM scores of 81.0/77.1/79.5 on the zh/en/zh-hard test sets, respectively. Across other benchmarks, dots.tts also consistently demonstrates open-source state-of-the-art performance, exhibiting strong generation stability, voice cloning ability, and emotional expressiveness. For efficient inference, we further apply CFG-aware MeanFlow distillation, enabling low-latency speech generation with first-packet latencies of 85/54 ms in output streaming and dual-streaming modes, respectively. To facilitate reproducible research and practical deployment, we release the training and inference code, together with the pretrained, post-trained, and MeanFlow-distilled checkpoints, under the Apache 2.0 license.

Confidence-based loss weighting is usually avoided in generative models because it accelerates errors when the model is confidently wrong, but this intuition breaks down in supervised diffusion training. We introduce the Eisbach log-barrier, a parameter-free weight derived from the entropy of the DiT output's spatial energy distribution: high entropy damps the gradient, while low entropy preserves it. Applied to LoRA fine-tuning of Stable Audio 3 Medium on MusicCaps, it unexpectedly yields stronger thematic development, clearer acoustic differentiation, and higher textural diversity than unweighted training, the opposite of mode collapse. This works because in supervised diffusion the gradient direction is locked to ground truth, so confidence only scales the step size, and because temporal entropy downweights flat samples while preserving high-contrast ones. The result is an online, self-referential data curriculum that emerges purely from the forward pass, with analyzed noise-level dynamics and testable predictions.

Speech anonymization is commonly evaluated using averagecase metrics such as the equal error rate, which can hide large disparities in re-identification risks across individuals. In this paper, we conduct a large-scale per-speaker privacy analysis using a linkability-based metric under a worst-case scenario. Nearly 5,000 speakers are evaluated across multiple anonymization systems, attacker architectures, and conversation lengths. While linkability scores are highly polarized at the speaker level, the sets of easy to re-identify and hard to re-identify speakers vary substantially across configurations. We show that no single factor explains speaker vulnerability. Instead, the re-identification risk emerges from the interaction between the attacker, the anonymizer, and the amount of available speech. These results challenge the notion of intrinsic speaker-level privacy risks and emphasize the need for evaluation protocols that are explicitly conditioned on the attacker and anonymizer.

We introduce MMAE, a Massive Multitask Audio Editing benchmark, serving as the first comprehensive evaluation testbed designed for general-purpose instruction-based audio editing. Spurred by the shift toward intelligent creation, interactive editing has rapidly expanded from visual domains, pioneered by models like Nano-banana 2 for images and Gemini-Omni for video, into audio. However, the current evaluation infrastructure lags severely, remaining highly fragmented and restricted to specific subdomains or basic operations. Unlike existing benchmarks that are limited in scope, MMAE extends to a broad spectrum of real-world scenarios, encompassing 7 distinct audio modalities, including sound, speech, music, and their mixtures. Furthermore, we establish a comprehensive taxonomy spanning 6 levels of task complexity, from basic modifications to multi-hop reasoning and multi-round editing, 2 levels of granularity, and 8 distinct operation types. Meticulously curated through human-agent collaboration, MMAE comprises 2,000 high-fidelity samples paired with a pioneering rubric-based evaluation framework. By decomposing free-form tasks into 17,741 verifiable criteria, this robust rubric-based paradigm enables a precise, multi-dimensional assessment of both instruction following and context consistency. Our extensive evaluation of leading models reveals that current systems remain far from achieving reliable edits. Strikingly, the Exact Match Rate (EMR) consistently falls below 5% and plummets to an absolute 0% in complex, mixed-modality tasks, exposing critical bottlenecks in precise execution and structural robustness. We hope MMAE will serve as a catalyst for future advances in the intelligent creation community, providing a clear diagnostic roadmap and establishing a standardized, long-lasting evaluation paradigm for next-generation audio editing systems.

Speech Emotion Conversion (SEC) aims to transform the emotion of a source utterance into a target emotion while preserving content and speaker identity. SEC on in-the-wild data is challenging due to the non-parallel nature of training data and complex real-world acoustics. Existing fixed-duration approaches either struggle to shift the emotion effectively (high quality, low conversion) or degrade speech naturalness (low quality, high conversion). We propose TargetSEC, an embedding-driven latent diffusion framework that generates emotion-focused style embeddings conditioned on speaker identity and continuous emotion. Unlike methods that diffuse over spectrograms, TargetSEC operates in a compact latent space. Experiments on the MSP-Podcast dataset show that TargetSEC outperforms current non-duration baselines in conversion accuracy while maintaining high speech quality, and achieves performance comparable to duration-prediction systems without explicit temporal modeling.

Instruction-following audio language models (ALMs) can be augmented with explicit acoustic cues, yet it remains unclear whether such cues are used in a grounded way when the raw audio is already available. We study this question in speech emotion recognition (SER) by deriving six interpretable acoustic concept tokens from the standardised eGeMAPS paralinguistic feature set. These tokens summarise energy, pitch, dynamics, brightness, formants, and voice quality, and are appended to the textual prompt while the audio input is kept unchanged. Across the widely used FAU-Aibo and IEMOCAP benchmarks, aligned tokens improve unweighted average recall (UAR), whereas shuffled, conflicting, or corrupted tokens reduce performance relative to aligned tokens and shift confusions toward neutral. Importantly, predictions do not collapse under strong token perturbations, suggesting that the models are sensitive to the symbolic cue channel but remain partly anchored to the audio signal. We argue that token-only interventions provide a practical way to probe audio-grounded cue use, robustness, and interpretability in ALM-based affective computing.

Text-guided audio editing aims to modify the language-specified acoustic content while preserving edit-irrelevant source components. Existing training-free methods typically rely on inversion-based editing. While inversion-free editing is appealing as it decreases computational overhead and reconstruction errors, it remains largely unexplored for audio editing. The key challenge is to construct a source-to-target editing path through diffusion denoising dynamics. In this paper, we introduce DirectAudioEdit, the first attempt to develop a training-free and inversion-free method for audio editing. Experiments on music and event-level benchmarks across two backbones show that DirectAudioEdit reduces macro-averaged FAD and KL by 15.9% and 15.8% compared with DDPM inversion, while achieving up to 64.5% editing speedup.

In recent years, audio generation has made significant progress in tasks such as text-to-speech (TTS), text-to-audio (TTA) and text-to-music (TTM). However, generating long-form and controllable audio from complex audio scene descriptions remains a significant challenge, as such scenes often require coordinated speech, sound effects, music, songs, temporal structure, and post-production. In this work, we introduce \textbf{Audio-Oscar}, a multi-agent framework for generating audio from complex descriptions. Audio-Oscar coordinates a set of specialist agents, each responsible for a different aspect of the audio scene, including character modeling and voice design, speech generation, fine-grained timeline planning, model selection, non-speech generation, and audio post-production. Audio-Oscar further incorporates feedback-driven refinement. In addition, to address the lack of suitable benchmarks for evaluating audio generation from complex audio scene descriptions, we construct \textbf{ASG-Bench}, an Audio Scene Generation Benchmark containing both scene descriptions paired with reference audio and text-only scene descriptions. Each scene is annotated with target audio events and temporal statements to evaluate whether the generated audio faithfully realizes the required scene content and temporal structure. Experimental results show that Audio-Oscar can effectively generate audio that matches complex scene descriptions. Project samples are available at https://audiooscar.github.io/. Our code is available at https://github.com/ziye26/Audio-Oscar.

Whisper, a widely adopted ASR model, is known to suffer from hallucinations - coherent transcriptions generated for non-speech audio entirely disconnected from the input. We investigate whether hallucinations can be detected and mitigated through Whisper's internal representations. We extract audio encoder activations and evaluate two representation spaces: raw Whisper activations and Sparse AutoEncoder (SAE) latents. We show that both spaces encode linearly separable hallucination-related information, with discriminative power concentrated in a sparse feature subset and increasing toward deeper encoder layers. We propose two steering strategies: activation-space steering and SAE latent-space steering. SAE-based steering reduces hallucination rate from 72.63% to 14.11% for Whisper small and from 86.88% to 27.33% for Whisper large-v3 on the full non-speech test set, with small WER degradation on speech data, approaching the performance of fine-tuning-based methods.

Recent neural audio codec-based speech generation (CodecFake) produces highly realistic audio, posing a challenge to existing deepfake countermeasure models. While using codec resynthesized speech (CoRS) as proxy data improves performance, it often suffers from limited generalization. We propose Domain-Shift Feature Augmentation (DSFA), which simulates "in-the-wild" variations by transforming deterministic feature statistics into stochastic distributions during fine-tuning. To evaluate generalization, we further introduce Codec-based Speech Generation Extension Evaluation (CoSG ExtEval) dataset, a more challenging extension of the CoSG Eval (from CodecFake+) dataset, featuring 40 unseen generative models and long-form audio. Experimental results demonstrate that combining a post-trained SSL backbone with DSFA effectively narrows the proxy-to-wild domain gap. This approach achieves state-of-the-art performance across diverse CodecFake attacks in both CoSG Eval and CoSG ExtEval.

We present BiEAR, a human auditory-inspired adaptive binaural front-end for multi-speaker localisation and distance estimation. Inspired by medial olivocochlear (MOC) feedback in human hearing, BiEAR uses a neural controller to adaptively adjust the frequency selectivity of a binaural auditory filterbank during inference. This yields time-frequency adaptive representations for ears, enabling the model to respond to changing acoustic conditions. We evaluate BiEAR on multi-speaker localisation and distance estimation in anechoic and real-room environments. Results show that the adaptive front-end improves localisation accuracy and robustness to unseen speakers and rooms compared with commonly used fixed binaural front-ends. Visualisation and analysis of learned filter adaptations show that BiEAR emphasises informative frequency bands over time. These findings suggest that adaptive, biologically inspired binaural front-ends can improve machine hearing robustness in complex acoustic scenes.

Large audio language models (LALMs) extend large language models with an audio encoder and large-scale audio data. However, the scarcity of high-quality annotated audio data remains a fundamental bottleneck for scaling. Through probing signal detectability analysis, we identify fine-grained spectrotemporal perceptual weaknesses in a foundation LALM. To address these challenges, we propose Spectrotemporal Counting (SpectCount), a data-efficient fine-tuning approach based on fully synthetic audio signals generated on-the-fly, without relying on real-world audio, annotations, or pretrained generative models. SpectCount not only resolves the observed weaknesses but also improves performance on diverse auditory benchmarks spanning sound, music, and speech, unseen during fine-tuning. These results suggest that weakness-targeted synthetic signals provide a data-efficient path toward enhanced auditory understanding capabilities in LALMs.

Cross-lingual voice cloning aims to generate speech in a target language while preserving speaker identity from a source-language reference. This task is central to speech translation and is the focus of the IWSLT 2026 Cross-Lingual Voice Cloning track. A key challenge is maintaining intelligibility and naturalness in the presence of accent variation and domain-specific vocabulary. We build on a multilingual text-to-speech model, FishAudio-S2-Pro, and introduce language tag prompting to improve language control and reduce accent leakage. We further apply reinforcement learning (RL) fine-tuning for task adaptation and observe improvements in intelligibility. Finally, we propose a reference-conditioned lexical matching method that improves pronunciation of domain-specific terms when lexical overlap is present. Results show that language prompting provides the largest gains, while lexical matching yields consistent improvements on matched subsets.

Current Audio-Visual Speech Recognition (AVSR) models achieve near-perfect performance on the standard LRS3 benchmark, raising concerns of adaptive overfitting. To systematically assess true generalisability, we construct a highly controlled, unseen evaluation set subsampled from the massive MultiVSR dataset. Unlike standard out-of-distribution benchmarks, our subset strictly matches the acoustic, visual, and demographic distributions of the LRS3 test set. Evaluating five state-of-the-art architectures reveals a universal performance collapse, proving that current systems fail to generalise even under strictly aligned conditions. Through a fine-grained attribute analysis across seven factors, we isolate the specific drivers of this degradation. Furthermore, we uncover a profound lexical bias, expose distinct error patterns, and surprisingly reveal that audio-visual performance even lags behind audio-only settings. We release our matched test set for future benchmarking.

Understanding what generative models retain from training data remains challenging, with implications for copyright and privacy. Beyond verbatim reproduction, models can encode subtler traces of their training data that never surface in their outputs yet remain exploitable. We study this regime for Rectified Flows, which are increasingly used in deployed generative systems. We analyse the interpolation path $X_λ= (1-λ)X_0 + λX_1$ that defines the Rectified Flow training. We show that a gap exists between the reconstruction of train and test data that follows a bell-shaped curve over $λ$, wich accumulates during training, while the validation metrics remain stable. The signal has a maximum whose location we derive in closed form under Gaussian assumptions. We validate these predictions on both audio and images and show that the bell-shaped structure is universal, while the peak prediction holds when our assumptions are satisfied. As a proof of concept, we exploit this specific $λ$-resolved structure to perform a Membership Inference Attack, distinguishing members of the training set from non-members.

Autoregressive "language" models (LMs) trained on raw waveforms can be repurposed for lossless audio compression, but prior work is limited to 8-bit audio, leaving open whether such approaches work for practical settings (16/24-bit) and can compete with existing codecs. We benchmark LM-based compression on full-fidelity audio across diverse domains (music, speech, bioacoustics), sampling rates (16kHz-48kHz), and bit depths (8, 16, 24-bit). Standard sample-level tokenization becomes intractable at higher bit depths due to vocabulary size (65K for 16-bit; 16.7M for 24-bit). We propose Trilobyte, a byte-level tokenization schema for full resolution audio, improving vocabulary scaling from $O(2^{b})$ to $O(1)$ and enabling the first tractable 24-bit LM-based lossless compression. While LMs consistently outperform FLAC and yield state-of-the-art compression at 8-bit and 16-bit, we observe that compression gains become more modest as bit depth increases beyond 8-bit.

A promising approach for steering auditory attention in complex listening environments relies on Auditory Attention Decoding (AAD), which aim to identify the attended speech stream in a multiple speaker scenario from neural recordings. Entrainment-based AAD approaches, typically assume access to clean speech sources and electroencephalography (EEG) signals to exploit low-frequency correlations between the neural response and the attended stimulus. In this study, we propose CA-TCN, a Causal-Anticausal Temporal Convolutional Network that directly classifies the attended speaker. The proposed architecture integrates several best practices from convolutional neural networks in sequence processing tasks. Importantly, it explicitly aligns auditory stimuli and neural responses by employing separate causal and anticausal convolutions respectively, with distinct receptive fields operating in opposite temporal directions. Experimental results, obtained through comparisons with three baseline AAD models, demonstrated that CA-TCN consistently improved decoding accuracy across datasets and decision windows, with gains ranging from 0.5% to 3.2% for subject-independent models and from 0.8% to 2.9% for subject-specific models compared with the next best-performing model, AADNet. Moreover, these improvements were statistically significant in four of the six evaluated settings when comparing Minimum Expected Switch Duration distributions. Beyond accuracy, the model demonstrated spatial robustness across different conditions, as the EEG spatial filters exhibited stable patterns across datasets. Overall, this work introduces an accurate and unified AAD model that outperforms existing methods while considering practical benefits for online processing scenarios. These findings contribute to advancing the state of AAD and its applicability in real-world systems.

MOSS-Audio is a unified audio-language model for speech, environmental sound, and music understanding, supporting audio captioning, time-aware question answering, timestamped transcription, and audio-grounded reasoning. MOSS-Audio couples a dedicated audio encoder with a modality adapter and a large language model: the encoder produces 12.5 Hz temporal representations, the adapter projects them into the decoder space, and the decoder generates autoregressive text outputs. Two design choices are central to the system: DeepStack cross-layer feature injection, which exposes the decoder to acoustic information from multiple encoder depths, and time markers, which provide explicit temporal cues by inserting timestamp markers into the audio-token stream. At the data level, we design an event-preserving audio annotation pipeline that segments raw audio at coherent event boundaries, applies branch-specific annotation to speech, music, and general audio, and merges the results into unified captions for pretraining. The intermediate branch-specific captions are further retained to support the construction of task-oriented SFT data. The model is pretrained on large-scale audio-language data, with time-aware objectives incorporated to support temporal grounding, and then undergoes multi-stage post-training to enhance instruction following and audio-grounded reasoning. We release 4B and 8B variants in both Instruct and Thinking configurations. MOSS-Audio achieves strong performance across general audio understanding, speech captioning, ASR, and timestamped ASR, positioning it as a promising understanding foundation for future voice agents.

World models simulate environmental dynamics to enable agents to plan and reason about future states. While existing approaches have primarily focused on visual observations, real-world perception inherently involves multiple sensory modalities. Audio provides crucial spatial and temporal cues such as sound source localization and acoustic scene properties, yet its integration into world models remains relatively underexplored. Prior work has not established a commonly adopted formulation for audio-visual world modeling under low-level action control or clarified how to jointly capture physically grounded binaural audio and visual dynamics. This work presents a unified formulation of Audio-Visual World Models (AVWM), casting multimodal environment simulation as a partially observable Markov decision process with synchronized audio-visual observations. As a foundational step toward this problem, we construct AVW-4k, a controlled benchmark comprising 30 hours of binaural audio-visual trajectories with action annotations across 76 indoor environments. We propose AV-CDiT, an Audio-Visual Conditional Diffusion Transformer with a novel modality expert architecture that balances visual and auditory learning, optimized through a three-stage training strategy for effective multimodal integration. Extensive experiments on this benchmark demonstrate that AV-CDiT achieves high-fidelity multimodal prediction across visual and auditory modalities. Furthermore, we validate its practical utility in embodied navigation, demonstrating that AVWM improves a vision-language-model-guided agent in continuous audio-visual navigation.

Audio-Visual Speech Recognition (AVSR) enhances speech recognition robustness by leveraging visual cues, while real-world scenarios remain challenging due to viewpoint variation, audio distortion, and visual occlusion, which degrade modality quality and increase audio-visual asynchrony. In this paper, we propose a novel Modality-aware Multi-view Self-supervised representation framework for robust Audio-Visual Speech Recognition (M2S-AVSR). First, we introduce a multi-view representation learning encoder to learn view-invariant visual speech representations. Next, we employ a modality-aware module that explicitly models modality quality and cross-modal synchrony to perform fine-grained modality-aware fusion, enabling fine-grained visual information injection during decoding. In addition, we release AISHELL8-RealScene, a public multi-scenario, multi-view conversational audio-visual dataset recorded in real-world environments, and establish a speech recognition benchmark on it. Experiments on English and Mandarin benchmarks demonstrate the effectiveness of the proposed method under challenging conditions. On LRS3, M2S-AVSR achieves up to 29.4% relative improvement under viewpoint perturbation and visual degradation settings. Our method also achieves new state-of-the-art performance on the MISP2021-AVSR test set. On AISHELL8-RealScene, it achieves the best result in outdoor scenes. The proposed method and dataset provide useful support for future research on robust speech and multimodal tasks under realistic conditions.

Voice disorders affect a significant portion of the population, and the ability to diagnose them using automated, non-invasive techniques would represent a substantial advancement in healthcare, improving the quality of life of patients. Recent studies have demonstrated that artificial intelligence models, particularly Deep Neural Networks (DNNs), can effectively address this task. However, due to their complexity, the decision-making process of such models often remain opaque, limiting their trustworthiness in clinical contexts. This paper investigates an alternative approach based on Explainable AI (XAI), a field that aims to improve the interpretability of DNNs by providing different forms of explanations. Specifically, this works focuses on concept-based models such as Concept Bottleneck Model (CBM) and Concept Embedding Model (CEM) and how they can achieve performance comparable to traditional deep learning methods, while offering a more transparent and interpretable decision framework.