SEARCH · Searcharxiv

Search Searcharxiv

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 73 records · Page 4Linked to original sources

arXiv2026

A Large-Scale Probing Analysis of Speaker-Specific Attributes in Self-Supervised Speech Representations

Enhancing explainability in speech self-supervised learning (SSL) is important for understanding and effectively utilising speech SSL representations. This study conducts a large-scale layer-wise probing analysis of 11 speech SSL models, examining speaker identity together with acoustic, prosodic, and paralinguistic attributes. The results confirm a general hierarchy wherein initial layers encode fundamental acoustics and middle layers synthesise abstract traits. The consensus that final layers purely abstract linguistic content is challenged, as larger models unexpectedly recover speaker-discriminative information in their deep layers. Furthermore, the intermediate representations of speech SSL models are found to provide stronger probing performance for several attributes than specialised speaker embeddings. These insights give a systematic empirical characterisation of speaker-related information in SSL models, providing guidelines for selecting representations for downstream tasks.

eess.AS↗

arXiv2026

Ensemble-Guided Distillation for Compact and Robust Acoustic Scene Classification on Edge Devices

We present a compact, quantization-ready acoustic scene classification (ASC) framework that couples an efficient student network with a learned teacher ensemble and knowledge distillation. The student backbone uses stacked depthwise-separable "expand-depthwise-project" blocks with global response normalization to stabilize training and improve robustness to device and noise variability, while a global pooling head yields class logits for efficient edge inference. To inject richer inductive bias, we assemble a diverse set of teacher models and learn two complementary fusion heads: z1, which predicts per-teacher mixture weights using a student-style backbone, and z2, a lightweight MLP that performs per-class logit fusion. The student is distilled from the ensemble via temperature-scaled soft targets combined with hard labels, enabling it to approximate the ensemble's decision geometry with a single compact model. Evaluated on the TAU Urban Acoustic Scenes 2022 Mobile benchmark, our approach achieves state-of-the-art (SOTA) results on the TAU dataset under matched edge-deployment constraints, demonstrating strong performance and practicality for mobile ASC.

cs.SD↗

arXiv2026

HoliTok: A Continuous Holistic Tokenization with Robust Dual Capabilities of Speech Generation and Understanding

Unified speech foundation models require a holistic tokenization space that is both learnable by language models and decodable into high-quality waveforms. Existing speech tokenizers, however, often fail to satisfy these requirements simultaneously, leading to increased architectural complexity and more involved training designs. We propose HoliTok, a continuous Holistic speech Tokenization model designed for unified generation-understanding modeling. HoliTok encodes 48~kHz speech into a compact 25~Hz sequence of 128-dimensional latents. It is trained with a progressive strategy that jointly preserves signal-level fidelity, incorporates semantic information, and maintains strong latent learnability. Based on this tokenization, we build a unified AR+DiT model for speech synthesis and recognition, where the same latent sequence supports both generation-specific and unified generation-understanding tasks. Experiments show that HoliTok achieves competitive reconstruction fidelity, improves generative learnability for high-quality and controllable synthesis, and, among the evaluated representations, is the only one that operates robustly in our unified generation-understanding architecture without additional optimization tricks. These results suggest that HoliTok serves as an effective speech tokenizer and a foundational representation interface for unified spoken language modeling. The code is available at: https://github.com/bovod-sjtu/HoliTok.

cs.SD↗

arXiv2026

Where Does the Sound Go? Tracing Acoustic Information Loss in Audio-Conditioned LLMs

Audio-conditioned language models often underuse acoustic cues such as prosody, emotion, and non-speech sounds, raising the question of whether ASR-supervised frontends discard this information before it reaches the LM. We test whether the frontend is responsible by comparing Whisper-Tiny and Whisper-Small with EnCodec, DAC-VAE, and WavTokenizer in a shared Qwen3.5-4B audio-LM pipeline on ASR, emotion recognition, and sound captioning. Encoder replacement alone does not resolve this underuse: Whisper variants remain strongest overall, including on emotion and environmental sound captioning. To localize the failure, we trace task-relevant information through the encoder, projector, LM layers, and LM head. Linear probes and geometric analyses show that discriminative acoustic structure remains recoverable at the final LM layer, even when MCQA accuracy trails probe accuracy by up to 83 points. Because the answer format and decoding procedure are controlled, this task-dependent gap points to content-specific readout failure rather than generic format bias. LogitLens analyses and a targeted LM head intervention support the conclusion that acoustic underuse is not explained solely by encoder-side information loss and that readout alignment can be a dominant bottleneck.

cs.SD↗

arXiv2026

Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts

Omni-modal large language models (OLLMs) jointly process vision, audio, and text, yet their modality bias under cross-modal conflict remains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality: perceptual signals (e.g., a photograph or recording of a dog) and propositional signals (e.g., the declarative claim "this is a dog"), such that any measured modality bias is inherently confounded with evidence-form bias, precluding clean attribution to either source. To address this, we introduce Tri-PvP, an 8,000-sample tri-modal conflict benchmark crossing vision, audio, and text, where vision and audio each take perceptual or propositional form. Evaluating five OLLMs, we find robust visual bias across most models and evidence-type conditions. Crucially, we reveal a systematic asymmetry in evidence-form bias: models exhibit a stronger bias toward perceptual signal in vision but propositional in audio. Further analyses via layer-wise linear probing and contrastive decoding reveal that modality bias is already linearly decodable from early representation layers and can only be partially mitigated, calling for mitigation strategies beyond surface-level interventions.

cs.CL↗

arXiv2026

SETEAB: Multiscale approach with Squeeze-and-Excitation Temporal Enhanced Aware Block for Speech Emotion Recognition

This paper proposes a novel lightweight multiscale architecture for speech emotion recognition (SER) with three key innovations. First, a depthwise convolution-based subsampling module is introduced to reduce model size and computation while preserving salient emotional cues. Second, a Squeeze-and-Excitation block is integrated to enhance channel-wise recalibration and improve representation robustness. Third, a new Temporal Enhanced Aware Block is designed to strengthen temporal dependency modeling and produce more discriminative emotion-aware features. The proposed model is explicitly designed to jointly improve compactness, recognition performance, and generalizability. Experiments on benchmark SER datasets show that our method achieves higher accuracy with reduced computational complexity, while also delivering stronger cross-corpus performance than most recent advanced networks for SER.

eess.AS↗

arXiv2026

CUHK-EE Systems for the vTAD Challenge at NCMMSC 2025

This paper presents the Voice Timbre Attribute Detection (vTAD) systems developed by the Digital Signal Processing & Speech Technology Laboratory (DSP&STL) of the Department of Electronic Engineering (EE) at The Chinese University of Hong Kong (CUHK) for the 20th National Conference on Human-Computer Speech Communication (NCMMSC 2025) vTAD Challenge. The proposed systems leverage WavLM-Large embeddings with attentive statistical pooling (ASTP) to extract robust speaker representations, followed by two variants of Diff-Net, i.e., Feed-Forward Neural Network (FFN) and Squeeze-and-Excitation-enhanced Residual FFN (SE-ResFFN), to compare timbre attribute intensities between utterance pairs. Experimental results demonstrate that the WavLM-Large+FFN system generalises better to unseen speakers, achieving 77.96% accuracy and 21.79% equal error rate (EER), while the WavLM-Large+SE-ResFFN model excels in the 'Seen' setting with 94.42% accuracy and 5.49% EER. These findings highlight a trade-off between model complexity and generalisation, and underscore the importance of architectural choices in fine-grained speaker modelling. Our analysis also reveals the impact of speaker identity, annotation subjectivity, and data imbalance on system performance, pointing to future directions for improving robustness and fairness in timbre attribute detection.

eess.AS↗

arXiv2026

AudioFuse: Unified Spectral-Temporal Learning via a Hybrid ViT-1D CNN Architecture for Robust Phonocardiogram Classification

Biomedical audio signals, such as phonocardiograms (PCG), are inherently rhythmic and contain diagnostic information in both their spectral (tonal) and temporal domains. Standard 2D spectrograms provide rich spectral features but compromise the phase information and temporal precision of the 1D waveform. We propose AudioFuse, an architecture that simultaneously learns from both complementary representations to classify PCGs. To mitigate the overfitting risk common in fusion models, we integrate a custom, wide-and-shallow Vision Transformer (ViT) for spectrograms with a shallow 1D CNN for raw waveforms. On the PhysioNet 2016 dataset, AudioFuse achieves a state-of-the-art competitive ROC-AUC of 0.8608 when trained from scratch, outperforming its spectrogram (0.8066) and waveform (0.8223) baselines. Moreover, it demonstrates superior robustness to domain shift on the challenging PASCAL dataset, maintaining an ROC-AUC of 0.7181 while the spectrogram baseline collapses (0.4873). Fusing complementary representations thus provides a strong inductive bias, enabling the creation of efficient, generalizable classifiers without requiring large-scale pre-training. The complete source code for our baseline and AudioFuse models is publicly available on GitHub at: https://github.com/Saiful185/AudioFuse.

eess.AS↗

arXiv2026

An Investigation of Reprogramming for Cross-Language Adaptation in Speaker Verification Systems

Language mismatch is among the most common and challenging domain mismatches in deploying speaker verification (SV) systems. Adversarial reprogramming has shown promising results in cross-language adaptation for SV. The reprogramming is implemented by padding learnable parameters on the two sides of input speech signals. In this paper, we investigate the relationship between the number of padded parameters and the performance of the reprogrammed models. Sufficient experiments are conducted with different scales of SV models and datasets. The results demonstrate that reprogramming consistently improves the performance of cross-language SV, while the improvement is saturated or even degraded when using larger padding lengths. The performance is mainly determined by the capacity of the original SV models instead of the number of padded parameters. The SV models with larger scales have higher upper bounds in performance and can endure longer padding without performance degradation.

eess.AS↗

arXiv2026

BFA: Real-time Multilingual Text-to-speech Forced Alignment

We present Bournemouth Forced Aligner (BFA), a system that combines a Contextless Universal Phoneme Encoder (CUPE) with a connectionist temporal classification (CTC)based decoder. BFA introduces explicit modelling of inter-phoneme gaps and silences and hierarchical decoding strategies, enabling fine-grained boundary prediction. Evaluations on TIMIT and Buckeye corpora show that BFA achieves competitive recall relative to Montreal Forced Aligner at relaxed tolerance levels, while predicting both onset and offset boundaries for richer temporal structure. BFA processes speech up to 240x faster than MFA, enabling faster than real-time alignment. This combination of speed and silence-aware alignment opens opportunities for interactive speech applications previously constrained by slow aligners.

eess.AS↗

arXiv2026

UniVoice: Unifying Autoregressive ASR and Flow-Matching based TTS with Large Language Models

Large language models (LLMs) have demonstrated promising performance in both automatic speech recognition (ASR) and text-to-speech (TTS) systems, gradually becoming the mainstream approach. However, most current approaches address these tasks separately rather than through a unified framework. This work aims to integrate these two tasks into one unified model. Although discrete speech tokenization enables joint modeling, its inherent information loss limits performance in both recognition and generation. In this work, we present UniVoice, a unified LLM framework through continuous representations that seamlessly integrates speech recognition and synthesis within a single model. Our approach combines the strengths of autoregressive modeling for speech recognition with flow matching for high-quality generation. To mitigate the inherent divergence between autoregressive and flow-matching models, we further design a dual attention mechanism, which switches between a causal mask for recognition and a bidirectional attention mask for synthesis. Furthermore, the proposed text-prefix-conditioned speech infilling method enables high-fidelity zero-shot voice cloning. Experimental results demonstrate that our method can achieve or exceed current single-task modeling methods in both ASR and zero-shot TTS tasks. This work explores new possibilities for end-to-end speech understanding and generation. Code is available at https://github.com/gwh22/UniVoice.

eess.AS↗

arXiv2026

SNAP: Speaker Nulling for Artifact Projection in Speech Deepfake Detection

Recent advancements in text-to-speech technologies enable generating high-fidelity synthetic speech nearly indistinguishable from real human voices. While recent studies show the efficacy of self-supervised learning-based speech encoders for deepfake detection, these models struggle to generalize across unseen speakers. Our quantitative analysis suggests these encoder representations are substantially influenced by speaker information, causing detectors to exploit speaker-specific correlations rather than artifact-related cues. We call this phenomenon speaker entanglement. To mitigate this reliance, we introduce SNAP, a speaker-nulling framework. We estimate a speaker subspace and apply orthogonal projection to suppress speaker-dependent components, isolating synthesis artifacts within the residual features. By reducing speaker entanglement, SNAP encourages detectors to focus on artifact-related patterns, leading to state-of-the-art performance.

cs.SD↗

arXiv2026

AudioKV: KV Cache Eviction in Efficient Large Audio Language Models

Large Audio-Language Models (LALMs) have set new benchmarks in speech processing, yet their deployment is hindered by the memory footprint of the Key-Value (KV) cache during long-context inference. While general KV cache compression techniques excel in LLMs, they often fail in the audio domain by overlooking the intrinsic temporal continuity of acoustic signals. To bridge this gap, we propose AudioKV, a novel framework that robustly prioritizes audio-critical attention heads through a hardware-friendly semantic-acoustic alignment mechanism. Specifically, we identify these modality-specialized heads by analyzing attention scores in ASR tasks and dynamically allocate KV cache budgets preferentially to them. Furthermore, we introduce Spectral Score Smoothing (SSS), an FFT-based global filtering strategy designed to suppress high-frequency noise and recover smooth global trends from importance scores, ensuring more balanced token selection with unprecedented precision. Extensive evaluations across multiple LALMs, including Qwen and Gemma series, demonstrate that AudioKV significantly outperforms baselines while enhancing computational efficiency. Notably, at a 40% compression ratio, AudioKV maintains near-full accuracy on Qwen3-Omni-30B with only a 0.45% drop, whereas traditional methods suffer from catastrophic performance degradation and repetition. Our code will be released after acceptance.

cs.SD↗

arXiv2026

Who Wins the Conflict? Mechanistic Interpretability of Text Bias in Audio LLMs

While Audio Large Language Models (Audio LLMs) excel at multimodal understanding, they suffer from text dominance, a bias where models favor text over acoustic evidence, potentially leading to hallucinated responses. However, the internal mechanisms underlying how these models behave when audio and textual inputs contradict each other remain unexplored. In this work, we present the first mechanistic analysis of this phenomenon by tracing the propagation of internal representations across layers. Our investigation reveals three key findings: (i) text dominance is consistently observed across models; (ii) while text and audio rely on functionally distinct pathways, they ultimately converge into a shared semantic space in late layers; and (iii) the text pathway does not erase audio information, but rather actively suppresses intact audio representations. Building on these insights, we leverage back-patching, a training-free intervention that routes late-layer audio activations back into earlier layers. This amplifies the audio representations, enabling them to overcome textual suppression. Our evaluation shows that back-patching consistently reduces text dominance, demonstrating a mechanistic route to mitigating text dominance under conflict.

cs.SD↗

arXiv2026

SCAPES: Semantically Conditioned Autoregressive Prior for Environmental Sounds

As generative audio models grow in complexity, the computational and ecological costs of synthesizing everyday sounds have become increasingly prohibitive, often requiring industrial-scale resources and massive datasets. In this paper, we present SCAPES: a Semantically Conditioned Autoregressive Prior for Environmental Sounds. SCAPES is a lightweight, resource-efficient generative model designed to synthesize high-fidelity environmental textures through high-level semantic control. By operating on the continuous latent manifold of a neural audio codec, our approach bypasses the rigid structural constraints inherent to discrete tokenization. We propose a segmentation strategy that decomposes audio into overlapping segments, enabling a Continuous Normalizing Flow (CNF) to model the evolution of latent trajectories using Flow Matching. Our experiments demonstrate that a 36-million parameter instance of SCAPES can be trained on limited, uncurated datasets using a single consumer-grade GPU. Notably, convergence is achieved after training for approximately twice the source audio duration, yielding high-fidelity outputs with robust long-term stability and semantic consistency. Furthermore, we showcase the model's capacity for smooth semantic interpolation, providing a flexible and accessible tool for open research and creative sound design. Code, pretrained weights, audio examples, and an interactive demo are publicly available on our project page https://cordutie.github.io/projects/scapes.html

cs.SD↗

arXiv2026

Tracing Audio Grounding and Answer Selection in Audio LLMs

Audio Large Language Models (Audio LLMs) have advanced in audio understanding, yet they can still predict the answer by reasoning from textual cues or linguistic priors rather than the provided audio. A common remedy is to train models on data whose answers cannot be inferred from text alone. This approach can improve performance, but what changes within the model remains unclear. In this paper, we ask what must happen inside the model for the audio to actually determine the answer. Our findings are threefold. (1) Replacing the audio with silence or unrelated audio causes substantially larger performance degradation in the trained model than in the pretrained model. (2) Acoustic information most strongly shapes the model's representations of the answer choices in early-to-middle layers, while training mainly increases the influence of audio information on the final prediction in middle-to-late layers. (3) The weights learned during training have their largest impact in specific layer bands. Together, these results provide a mechanistic account of how training strengthens the use of acoustic evidence in Audio LLMs.

cs.CL↗

arXiv2026

Harmonica: Accurate and Lightweight Instrument-Agnostic Music Transcription

This paper introduces Harmonica, a family of instrument-agnostic music transcription models built around multi-depth harmonic convolution. At each model scale, Harmonica achieves the best performance among the evaluated models: the x-large model attains state-of-the-art performance in instrument-agnostic transcription, while the medium variant offers competitive accuracy with faster inference than all baselines. Pushing the limit of computational efficiency, the nano variant has only 26.3K parameters and runs at 1,622.5 times real time, yet achieves a frame F1 of 0.796 on the development set, 14.6 percentage points higher than Basic Pitch. We further demonstrate that multi-depth harmonic convolution effectively exploits harmonic information to benefit transcription performance through comparative experiments with existing harmonic aggregation methods, including harmonic stacking, harmonic attention, single-depth harmonic convolution, and the HD-Conv layer.

cs.SD↗

arXiv2026

What Selects, What Reconstructs: Repairing Exemplar-Based Complex-Spectrum Separation

Exemplar methods separate a mixture by picking one learned spectrum per source and deforming it until it explains the observation, making one deformation class both reconstructor and selector. We show that the second role is empty as soon as the class can interpolate: the rule then ranks candidates on its regulariser, a choice made before the data, and the estimates sum back to the mixture whichever candidate wins. The condition is a parameter count, so the diagnosis runs before any experiment. On free per-bin deformation of complex spectra it explains the observed pathologies at once: a criterion that ranks candidates by their loudness, and half an output that is a mask on the mixture rather than an exemplar. The same theorem prescribes the repair, a selection class poorer than the reconstruction class: one complex gain and one pure delay rank the candidates, and a local combination of the best-aligned atoms, fitted jointly in closed form, rebuilds them. On MUSDB18 against the exact ceiling of the masking class, the distance between the criterion and an oracle inside its own candidate pool falls under the rigid selector from 6.2-7.7 to 0.5-2.8 dB, though only 0.9-1.2 dB of that reaches the output, and the per-frame latency of the deployed rule by a factor of 47 to 806. One lock remains, quantified: atoms are scored against the mixture, so the score carries a term for the other source that absorbs the capacity the reconstruction class gains, leaving the output 10.0 dB under the ceiling. Ranking hypotheses by the residual of a fit free enough to interpolate ranks them on the regulariser alone.

eess.SP↗