Generative video-to-audio (V2A) models produce highly plausible soundtracks, but it remains unclear whether they capture the underlying physical processes. Existing evaluations emphasize perceptual realism and overlook physical correctness under controlled interventions. In this paper, we introduce FlatSounds, a benchmark that audits the physical reasoning of V2A models through: 1) controlled counterfactual pairs in which a single physical factor is varied, and 2) single-video pattern tests that probe internal consistency and directional trends. These settings test whether the generated audio correctly reflects specific physical properties and timings. Our evaluation of state-of-the-art models reveals a consistent trade-off: models rely more on text captions than the visual stream to infer physics and semantics. Captions generally improve physical and semantic accuracy, but paradoxically degrade temporal alignment. Our results highlight the need to move beyond audio quality toward learning physical processes directly from pixels. Finally, we find that our physics-based metrics correlate strongly with human preference tests on our own data. Project webpage: https://research.nvidia.com/labs/cosmos-lab/flatsounds/

Large Audio Language Models (LALMs) expand jailbreak risks from token-level prompting to the full speech perception-to-reasoning pipeline, where unsafe behavior can be induced through semantics, acoustic style, signal artifacts, or internal representations. Existing work studies these risks under heterogeneous threat models and evaluation protocols, making it difficult to compare attack practicality or defense utility. This paper provides a unified taxonomy and a controlled empirical evaluation of LALM jailbreak attacks and defenses. We organize prior work into semantic, acoustic, signal, and embedding-layer attacks; guard-based, training-free, and training-based defenses; and cross-modal, audio-native, and interactive benchmarks. We then evaluate representative attacks and defenses across ten open-source LALMs, measuring not only attack success rate but also benign refusal and latency. Our results show that Acoustic Best-of-N reveals strong worst-case audio-space vulnerabilities, Narrative Framing is an effective low-latency semantic threat, and current defenses trade robustness against benign usability. These findings support cost- and utility-aware evaluation as a necessary complement to success-rate-only LALM safety benchmarks.

AI-driven respiratory sound classification (RSC) is promising for automated pulmonary disease detection, yet multi-site deployment is hindered by inter-stethoscope variability. We introduce a federated domain generalization (FedDG) formulation for RSC under stethoscope-induced device shifts, where clients use heterogeneous devices and the model is evaluated on unseen devices. Our empirical analysis shows that stethoscope-induced style and disease-specific content are tightly entangled, making deterministic style removal unreliable. In response, we propose a causality-inspired multimodal FedDG framework that combines: (i) a causality-inspired device style intervention network that performs content-preserving style perturbations, (ii) counterfactual text augmentation that neutralizes metadata shortcuts, and (iii) gradient alignment that facilitates device-invariant representations across clients. Built on a multimodal language-audio pretraining model, it outperforms conventional data augmentation and federated learning baselines in leave-one-device-out validation on ICBHI and SPRSound datasets. Code will be released upon publication.

Contrastive Language-Audio Pretraining (CLAP) models are widely used for audio understanding and support modality-agnostic condition swapping in many zero-shot applications. However, their performance is heavily affected by the modality gap between audio and text embeddings. Existing explanations mainly attribute this gap to the cone effect, treating it as a shift between mean embeddings, yet correcting the mean alone yields only limited improvements. Alternative hypotheses, such as information imbalance and dimensionality collapse, have also been proposed, but they remain insufficiently verified and have not been thoroughly studied in the audio domain. Meanwhile, several works attempt to decompose multimodal contrastive embeddings into interpretable concepts, but none explicitly analyze the modality gap from the perspective of concept decomposition. In this work, we introduce COMET (Concept space Organization and Modality gap Explanation with PLS-SVD Transformation), a novel partial least squares singular value decomposition (PLS-SVD) framework for CLAP that unveils a broader perspective of the modality gap. Our framework reveals that only a small, interpretable subset of axes, which captures shared concepts, contributes substantially to similarity computation, and that the mean component represents only partially the modality gap. Building on this insight, we propose a simple spectral truncation method that mitigates the modality gap in a training-free manner. The method enables zero-shot audio captioning with condition swapping to approach fully supervised performance, without requiring large auxiliary memory banks or expensive computation. At the same time, it achieves substantial embedding dimensionality reduction while preserving strong performance on retrieval and audio captioning tasks.

While LLM-based Automatic Speech Recognition (ASR) achieves high accuracy, its speed is limited by sequential autoregressive decoding. Diffusion Language Models (DLMs) offer a parallel alternative, yet their decoding strategies remain under-explored in ASR contexts. This paper analyzes three decoding schemes for DLM-based ASR: fixed-number, static confidence threshold, and dynamic confidence threshold. We propose measuring round-wise accuracy using Negative Log-Likelihood-based uncertainty as a proxy for decoding progress. Our results show that both threshold-based strategies significantly outperform fixed-number schemes in accuracy and speed. We attribute this to a property unique to ASR: most tokens reach high confidence early, allowing reliable ones to be harvested aggressively while leaving only difficult tokens for later rounds. Notably, the static-threshold strategy matches the accuracy of autoregressive decoding while offering superior efficiency.

Medical audio data is difficult to collect due to privacy regulations and high annotation costs arising from domain expertise. Thus, existing benchmarks tend to underrepresent complex medical audio scenarios. To address this challenge, we present MedMosaic, a medical audio question-answering dataset designed to benchmark language and audio reasoning models under realistic clinical constraints. MedMosaic features a diverse range of medical audio types, including condition-related physiological sounds, carefully constructed synthetic voices to mimic speech with artifacts as well as real short and long length clinical conversations to model varying context lengths. The dataset also features a total of 46,701 question-answer pairs, spanning categories such as multiple-choice, sequential multi-turn, and open-ended question-answers, enabling systematic evaluation of multi-hop reasoning and answer generation capabilities. Benchmarking 13 audio and multimodal reasoning models reveals that reasoning remains challenging for all evaluated systems, with substantial performance variation across question types. In particular, even state-of-the-art model like Gemini-2.5-pro can only achieve 68.1% accuracy approximately. These findings underscore persistent limitations in medical reasoning and highlight the need for more robust, domain-specific multimodal reasoning models. A sample of benchmark data is available here: https://shorturl.at/Lyp33

Large Audio Language Models (LALMs) still struggle in complex acoustic scenes because they often fail to preserve task-relevant acoustic evidence before reasoning begins. We identify this error pattern as the evidence bottleneck: state-of-the-art systems show larger deficits in acoustic evidence extraction than in downstream reasoning, suggesting that upstream perception is often the limiting factor. To address this problem, we propose EvA (Evidence-First Audio), a dual-path architecture that enhances acoustic evidence preservation through hierarchical aggregation and non-compressive, time-aligned fusion. We also build EvA-Perception, a large-scale training set with about 54K event-ordered captions and 500K evidence-grounded QA pairs. Under a unified zero-shot protocol, EvA achieves the best open-source \emph{Perception} results on MMAU, MMAR, and MMSU, with the largest gains on perception-heavy splits. Human evaluation on open-ended captioning further shows improved fine-grained acoustic coverage and caption quality. These results support the evidence-first hypothesis: stronger audio understanding depends on preserving acoustic evidence before reasoning. Project can be found at https://satsuki2486441738.github.io/EvA/.

Current audio-visual large language models (AV-LLMs) are predominantly restricted to 2D perception, relying on RGB video and monaural audio. This design choice introduces a fundamental dimensionality mismatch that precludes reliable source localization and spatial reasoning in complex 3D environments. We address this limitation by presenting JAEGER, a framework that extends AV-LLMs to 3D space, to enable joint spatial grounding and reasoning through the integration of RGB-D observations and multi-channel first-order ambisonics. A core contribution of our work is the neural intensity vector (Neural IV), a learned spatial audio representation that encodes robust directional cues to enhance direction-of-arrival estimation, even in adverse acoustic scenarios with overlapping sources. To facilitate large-scale training and systematic evaluation, we propose SpatialSceneQA, a benchmark of 61k instruction-tuning samples curated from simulated physical environments. Extensive experiments demonstrate that our approach consistently surpasses 2D-centric baselines across diverse spatial perception and reasoning tasks, underscoring the necessity of explicit 3D modelling for advancing AI in physical environments. Our source code, pre-trained model checkpoints, and datasets are available at https://github.com/liuzhan22/JAEGER.

Existing mainstream video customization methods focus on generating identity-consistent videos based on given reference images and textual prompts. Benefiting from the rapid advancement of joint audio-video generation, this paper proposes a more compelling new task: sync audio-video customization, which aims to synchronously customize both video identity and audio timbre. Specifically, given a reference image $I^{r}$ and a reference audio $A^{r}$, this novel task requires generating videos that maintain the identity of the reference image while imitating the timbre of the reference audio, with spoken content freely specifiable through user-provided textual prompts. To this end, we propose OmniCustom, a powerful DiT-based audio-video customization framework that can synthesize a video following reference image identity, audio timbre, and text prompts all at once in a zero-shot manner. Our framework is built on three key contributions. First, identity and audio timbre control are achieved through separate reference identity and audio LoRA modules that operate through self-attention layers within the base audio-video generation model. Second, we introduce a contrastive learning objective alongside the standard flow matching objective. It uses predicted flows conditioned on reference inputs as positive examples and those without reference conditions as negative examples, thereby enhancing the model ability to preserve identity and timbre. Third, we train OmniCustom on our constructed large-scale, high-quality audio-visual human dataset. Extensive experiments demonstrate that OmniCustom outperforms existing methods in generating audio-video content with consistent identity and timbre fidelity. Project page: https://omnicustom-project.github.io/page/.

Audio chaptering, the task of segmenting long-form audio into coherent sections, is increasingly important for navigating podcasts, lectures, and videos. Despite its relevance, research remains limited and text-based, leaving key questions unresolved about leveraging audio information, handling ASR errors, and transcript-free evaluation. We address these gaps through three contributions: (1) a systematic comparison between text-based models with acoustic features, a novel audio-only architecture (AudioSeg) operating on learned audio representations, and multimodal LLMs; (2) empirical analysis of factors affecting performance, including transcript quality, acoustic features, duration, and speaker composition; and (3) formalized evaluation protocols contrasting transcript-dependent text-space protocols with transcript-invariant time-space protocols. Our experiments on YTSeg reveal that AudioSeg substantially outperforms text-based approaches, pauses provide the largest acoustic gains, and MLLMs remain limited by context length and weak instruction following, yet MLLMs are promising on shorter audio.

Recent advancements in song generation have shown promising results in generating songs from lyrics and/or global text prompts. However, most existing systems lack the ability to model the temporally varying attributes of songs, limiting fine-grained control over musical structure and dynamics. In this paper, we propose SegTune, a non-autoregressive framework for structured and controllable song generation. SegTune enables segment-level control by allowing users or large language models to specify local musical descriptions aligned to song sections.The segmental prompts are injected into the model by temporally broadcasting them to corresponding time windows, while global prompts influence the whole song to ensure stylistic coherence. To obtain accurate segment durations and enable precise lyric-to-music alignment, we introduce an LLM-based duration predictor that autoregressively generates sentence-level timestamped lyrics in LRC format. We further construct a large-scale data pipeline for collecting high-quality songs with aligned lyrics and prompts, and propose new evaluation metrics to assess segment-level alignment and vocal attribute consistency. Experimental results show that SegTune achieves superior controllability and musical coherence compared to existing baselines. See https://cai525.github.io/SegTune_demo for demos of our work.

Emotions conveyed through voice and face shape engagement and context in human AI interaction. Despite rapid progress in omni modal large language models, the holistic evaluation of emotional reasoning with audiovisual cues remains limited. To address this gap, we introduce AV EMO Reasoning, a benchmark designed to systematically assess emotional reasoning abilities in large language models. The framework uses a curated audiovisual corpus comprising synthetic single turn and multi turn dialogues and a real world subset, together with emotion perception and interaction reasoning metrics, to evaluate whether models can understand user emotions and produce appropriate responses. By releasing a systematic evaluation benchmark, AV EMO Reasoning offers a reproducible standard for evaluating emotion aware dialogue and advances toward more natural, adaptive human AI interaction.

Passive acoustic monitoring (PAM) systems generate continuous recordings spanning months, yet automated bioacoustic analysis of whale calls requires two separate annotation efforts: binary presence labels for classification and precise temporal boundaries for localization. A binary label for a multi-minute recording can be assigned in seconds, but timestamping every call within it requires hours of expert effort. Providing both is infeasible at operational scale. We present DSMIL-LocNet, a weakly supervised multiple instance learning (MIL) framework that performs both classification and temporal localization using only recording-level presence/absence labels. Our dual-stream architecture integrates spectral and temporal features to process recordings of 2--30 minutes without the temporal compression that degrades existing CNN methods on long inputs. On the AcousticTrends BlueFinLibrary, DSMIL-LocNet achieves F1 scores of 0.88--0.91 on recordings of 300--1800s, where fully supervised CNN baselines degrade to 0.19--0.64. It also provides temporal localization that these baselines cannot produce without frame-level annotation. Code: https://github.com/Ragib-Amin-Nihal/DSMIL-Loc

Audio deepfake detection is well-studied as a binary problem, but partially manipulated speech, where a short synthesised segment is spliced into an otherwise genuine utterance, poses a harder and more realistic threat. Detecting such half-truth audio requires not only distinguishing it from real and fully fake speech, but also localising where the manipulation occurs. We present CAFNet, a 576k-parameter architecture that addresses both tasks jointly: it performs ternary classification (real, fully-fake, or half-truth) and regresses the temporal boundaries of the synthesised region in a single forward pass. CAFNet fuses Mel-Frequency Cepstral Coefficient (MFCC), Linear-Frequency Cepstral Coefficient (LFCC), and Chroma Short-Time Fourier Transform (Chroma-STFT) features through parallel depthwise-separable convolution branches with cross-attention, followed by a Bidirectional Long Short-Term Memory (BiLSTM) regression head for boundary prediction. On the combined Multi-Lingual Audio Deepfake Detection Corpus (MLADDC) T2+T3 test set, CAFNet achieves 92.71% accuracy and macro Area Under the Curve (AUC) of 0.9910, with boundary localisation Mean Absolute Error (MAE) of 0.075s and a median error of 0.052s. On binary detection, it achieves 96.76% accuracy and 3.20% Equal Error Rate (EER), outperforming fine-tuned XLS-R 300M (78.31%) and AST 87M (93.03%) at over 500 times fewer parameters. A cross-dataset study further shows that standard fine-tuning collapses cross-domain representations even under reduced backbone learning rates.

Recent Large Audio-Language Models (LALMs) have demonstrated promising abilities in understanding musical content. However, whether their responses are grounded in the correct temporal regions of the audio remains underexplored. This limitation is particularly critical for music understanding, where key information often occurs as temporally localized events, such as instrument entries and rhythmic transitions. To address this gap, we introduce MusTBENCH, a music-expert-validated benchmark designed to evaluate temporal grounding in LALMs through five temporally grounded question-answering tasks. To further improve temporal grounding in existing models, we propose MusT, a novel four-stage temporal optimization recipe spanning music encoder adaptation, LLM adaptation, LLM supervised fine-tuning, and RL-based optimization. Experiments on MusTBENCH show that existing LALMs struggle with precise temporal grounding, while MusT brings significant improvements over strong baselines. These results establish temporal grounding as a key missing capability in current LALMs and position MusTBENCH as a challenging benchmark for future research in temporally grounded music understanding.

We present ChildVox, a novel benchmark for characterizing the diverse acoustic signals through which children communicate. Specifically, ChildVox follows the full developmental trajectory from birth through school age, covering physiological sounds, non-linguistic vocalizations, canonical syllables, and spoken language. ChildVox integrates more than 20 sub-tasks across 17 child-centered audio and speech datasets, enabling systematic cross-corpus and cross-domain comparison. We evaluate a representative range of audio and speech foundation models, including self-supervised, ASR-oriented, and large audio-language models, on tasks including physiological sound classification, vocalization and canonical syllables modeling, and speech quality assessment and recognition. Benchmark results show that ChildVox provides a suite of high-performance models in recognizing a wide range of acoustic signals from children, supporting downstream applications such as characterizing children's language levels and tracking speech production with age.

Voice agents, artificial intelligence systems that conduct spoken conversations to complete tasks, are increasingly deployed across enterprise applications. However, no existing benchmark jointly addresses two core evaluation challenges: generating realistic simulated conversations, and measuring quality across the full scope of voice-specific failure modes. We present EVA-Bench, an end-to-end evaluation framework that addresses both. On the simulation side, EVA-Bench orchestrates bot-to-bot audio conversations over dynamic multi-turn dialogues, with automatic simulation validation that detects user simulator error and appropriately regenerates conversations before scoring. On the measurement side, EVA-Bench introduces two composite metrics: EVA-A (Accuracy), capturing task completion, faithfulness, and audio-level speech fidelity; and EVA-X (Experience), capturing conversation progression, spoken conciseness, and turn-taking timing. Both metrics apply to all major agent architectures, enabling direct cross-architecture comparison. EVA-Bench includes 213 scenarios across three enterprise domains, a controlled perturbation suite for accent and noise robustness, and pass@1, pass@k, pass^k measurements that distinguish peak from reliable capability. Across 12 systems spanning all three architectures, we find: (1) no system simultaneously exceeds 0.5 on both EVA-A pass@1 and EVA-X pass@1; (2) peak and reliable performance diverge substantially (median pass@k--pass^k gap of 0.44 on EVA-A); and (3) accent and noise perturbations expose substantial robustness gaps, with effects varying across architectures, systems, and metrics (mean $Δ$ up to 0.314). We release the full framework, evaluation suite, and benchmark data under an open-source license.

REPresentation Alignment (REPA) improves the training of generative flow models by aligning intermediate hidden states with pretrained teacher features, but its effectiveness in token-conditioned audio Flow Matching critically depends on the choice of supervised layers, which is typically made heuristically based on the depth. In this work, we introduce Attribution-Guided REPresentation Alignment (AG-REPA), a novel causal layer selection strategy for representation alignment in audio Flow Matching. Firstly, we find that layers that best store semantic/acoustic information (high teacher-space similarity) are not necessarily the layers that contribute most to the velocity field that drives generation, and we call it Store-Contribute Dissociation (SCD). To turn this insight into an actionable training guidance, we propose a forward-only gate ablation (FoG-A) that quantifies each layer's causal contribution via the induced change in the predicted velocity field, enabling sparse layer selection and adaptive weighting for alignment. Across unified speech and general-audio training (LibriSpeech + AudioSet) under different token-conditioning topologies, AG-REPA consistently outperforms REPA baselines. Overall, our results show that alignment is most effective when applied to the causally dominant layers that drive the velocity field, rather than to layers that are representationally rich but functionally passive.

Recent advances in Large Audio Language Models (LALMs) have extended Text-to-Speech (TTS) to interactive role-play scenarios, which demand high expressiveness and strict adherence to role-play instructions. However, existing models struggle to maintain stylistic consistency with character profiles and scene descriptions across multi-turn dialogues. A critical bottleneck is the lack of objective metrics for quantifying speaking style. To bridge this gap, we propose Mean Continuation Log-Probability (MCLP) as both an evaluation metric and a reward signal, validated on LALM-based Role-Play TTS (RP-TTS) tasks. MCLP leverages the in-context learning capability of pretrained LALMs to measure the likelihood of ground-truth speech tokens conditioned on a contextual history consisting of the transcript, generated speech, and repeated transcript, serving as a proxy for stylistic continuity. Furthermore, we employ MCLP as a reinforcement learning reward to enhance the style alignment between generated speech and role-play instructions. To support this task, we construct a large-scale RP-TTS dataset with rich scene and character annotations. Experiments demonstrate that MCLP is well aligned with human judgments of stylistic consistency and serves as an effective reward for improving RP-TTS, leading to consistent gains in both objective metrics and subjective evaluations. Our code is publicly available at https://github.com/y-ren16/MCLP.

We present a thorough analysis of the findings of the latest iteration of the Singing Voice Conversion Challenge, a scientific event aiming to compare and understand different voice conversion systems in a controlled environment. Compared to previous iterations which solely focused on converting the singer identity, this year we also focused on converting the singing style of the singer. To create a controlled environment and thorough evaluations, we developed a new challenge database, introduced two tasks, open-sourced baselines, and conducted large-scale crowd-sourced listening tests and objective evaluations. The challenge was run for two months and in total we evaluated 33 different systems. The results of the large-scale crowd-sourced listening test showed that top systems had comparable singer identity scores to ground truth samples. However, modeling the singing style and consequently achieving high naturalness still remains a challenge in this task, primarily due to the difficulty in modeling dynamic information in breathy, glissando, and vibrato singing styles. Further analyses of the challenge also discuss the limitations of both the traditional similarity test and the dynamic preference test in evaluating singing style similarity. Moreover, calculating Spearman's rank correlation coefficient shows that dependent objective metrics such as chroma-alignment and non-match metrics such as speaker embeddings are the most correlated to subjective scores, but are still not at a level where it could be considered as a true replacement for subjective scores.

Monaural multi-speaker automatic speech recognition (ASR) remains challenging due to data scarcity and the intrinsic difficulty of recognizing and attributing words to individual speakers, particularly in overlapping speech. Recent advances have driven the shift from cascade systems to end-to-end (E2E) architectures, which reduce error propagation and better exploit the synergy between speech content and speaker identity. Despite rapid progress in E2E multi-speaker ASR, the field lacks a comprehensive review of recent developments. This survey provides a systematic taxonomy of E2E neural approaches for multi-speaker ASR, highlighting recent advances and comparative analysis. Specifically, we analyze: (1) architectural paradigms (SIMO vs.~SISO) for pre-segmented audio, analyzing their distinct characteristics and trade-offs; (2) recent architectural and algorithmic improvements based on these two paradigms; (3) extensions to long-form speech, including segmentation strategy and speaker-consistent hypothesis stitching. Further, we (4) evaluate and compare methods across standard benchmarks. We conclude with a discussion of open challenges and future research directions towards building robust and scalable multi-speaker ASR.