Audio Representation Learning

Follow this topic Ask about this topic

Audio representation learning studies how to encode waveforms or spectrograms into features that preserve information useful for tasks such as speech recognition, sound classification, music retrieval, and generation. Common approaches include self-supervised masked prediction in wav2vec 2.0, HuBERT, and BEST-RQ; contrastive audio-language learning such as CLAP; and distillation or autoencoder-based continuous and discrete audio tokenizers. Researchers assess transfer with linear probes and fine-tuning on benchmarks including SUPERB, ESC-50, MARBLE, and X-ARES, while also probing which acoustic, semantic, spatial, or speaker attributes the features retain. A central tension is how to make representations broadly transferable and compact while preserving fine acoustic detail, disentangling nuisance factors, and supporting both understanding and high-fidelity generation.

Latest papers

211 papers30 papers in the last four weeks, up 25% on the four weeks before. 0.2% of all new papers.

May 6, 2026

Benchmarking LLMs on the Massive Sound Embedding Benchmark (MSEB)

The Massive Sound Embedding Benchmark (MSEB) has emerged as a standard for evaluating the functional breadth of audio models. While initial baselines focused on specialized encoders, the shift toward "audio-native" Large Language Models (LLMs) suggests a new paradigm where a single multimodal backbone may replace complex, task-specific pipelines. This paper provides a rigorous empirical evaluation of leading LLMs - including members from the Gemini and GPT families - across the eight core MSEB capabilities to assess their efficacy and audio-text parity. Our results indicate that while a significant modality gap persists regarding performance and robustness, the empirical evidence for an "optimal" modeling approach remains inconclusive. Ultimately, the choice between audionative and cascaded architectures depends heavily on specific use-case requirements and the underlying assumptions regarding latency, cost, and reasoning depth.

Cyril Allauzen, Tom Bagby, Georg Heigold +2

Google USA & Germany

Audio-Language Model Evaluation LLM Evaluation Audio Representation Learning Audio-Language Models

May 5, 2026

PHALAR: Phasors for Learned Musical Audio Representations

Stem retrieval, the task of matching missing stems to a given audio submix, is a key challenge currently limited by models that discard temporal information. We introduce PHALAR, a contrastive framework achieving a relative accuracy increase of up to

\approx 70\%

over the state-of-the-art while requiring

<50\%

of the parameters and a 7

\times

training speedup. By utilizing a Learned Spectral Pooling layer and a complex-valued head, PHALAR enforces pitch-equivariant and phase-equivariant biases. PHALAR establishes new retrieval state-of-the-art across MoisesDB, Slakh, and ChocoChorales, correlating significantly higher with human coherence judgment than semantic baselines. Finally, zero-shot beat tracking and linear chord probing confirm that PHALAR captures robust musical structures beyond the retrieval task.

Davide Marincione, Michele Mancusi, Giorgio Strano +4

Department of Computer Science, Sapienza University of Rome, Italy · Moises Systems, Inc. · Paradigma, Inc.

Audio Representation Learning Representation Learning Audio Understanding Music Information Retrieval Equivariant Representation Learning Music Understanding

May 3, 2026

Shao: Scaling Acoustic Token Language Models Toward High-Fidelity Music Generation

A common design pattern in high-quality music generation is to handle structure and fidelity in different representation spaces: a generator first models high-level structure, followed by diffusion-based or neural decoding stages that reconstruct fine details. In this work, we explore an alternative view: both may be progressively modeled within a single deep acoustic-token hierarchy. To study this, we build a 64-layer residual vector quantization (RVQ) acoustic representation and propose a two-stage coarse-to-fine generation framework. A backbone model first generates coarse acoustic tokens for the full track, and a super-resolution model then completes finer tokens within the same acoustic token space. The super-resolution stage works at full-track scale and refines tokens layer by layer while running in parallel over time, leading to a fixed 62-step inference process. To jointly improve lyric alignment and fine-detail reconstruction, we further introduce hybrid-attention training: the alignment objective uses causal attention, while layer-wise refinement uses full attention. A key finding is that text-vocal alignment can emerge within pure acoustic-token language modeling, without requiring a separate semantic token stage. Moreover, initializing the super-resolution model from the trained backbone significantly improves convergence and final quality. Taken together, our results suggest that high-quality music generation can be effectively pursued without separating structure and fidelity into heterogeneous representation spaces. Instead, both can be progressively modeled within a unified acoustic-token hierarchy, pointing toward a simpler and more unified path to high-quality music generation.

Jiafeng Liu, Yuanliang Dong, Hongjia Liu +8

1Central Conservatory of Music · 2Tsinghua University

Text-to-Music Generation Audio Representation Learning Residual VQ Music Generation

May 1, 2026

LASE: Language-Adversarial Speaker Encoding for Indic Cross-Script Identity Preservation

A speaker encoder used in multilingual voice cloning should treat the same speaker identically regardless of which script the audio was uttered in. Off-the-shelf encoders do not, and the failure is accent-conditional. On a 1043-pair Western-accented voice corpus across English, Hindi, Telugu, and Tamil, WavLM-base-plus-sv loses 0.082 absolute cosine similarity when the same voice changes script and ECAPA-TDNN loses 0.105. On a 1369-pair Indian-accented voice corpus, the gap shrinks to 0.006 (WavLM-SV) and 0.044 (ECAPA-TDNN). The leak is largest where it matters most for cross-script TTS: when a system projects a non-Indic-trained voice into Indic scripts. We present LASE (Language-Adversarial Speaker Encoder), a small projection head over frozen WavLM-base-plus trained with two losses: a supervised contrastive loss over voice identity, and a gradient-reversal cross-entropy against a 4-language classifier that pushes the embedding to be language-uninformative while remaining speaker-informative. Trained on 1118 quality-gated cross-script pairs synthesised from 8 commercial multilingual voices, LASE's residual gap is consistent with zero on both corpora (Delta = 0.013 Western, Delta = 0.026 Indian; both bootstrap 95% CIs include zero) and amplifies the cross-script-vs-floor margin 2.4-2.7x over both baselines. An ECAPA+GRL ablation shows the GRL objective improves either backbone but the WavLM choice contributes too. In synthetic multi-speaker diarisation, LASE matches ECAPA-TDNN on cross-script speaker recall (0.788 vs 0.789) with ~100x less training data. We release the r1 checkpoint, both corpora, and the bootstrap recipe.

Venkata Pushpak Teja Menta

Praxel Ventures

Adversarial Training Adversarial Representation Learning Audio Representation Learning Speaker Verification

Apr 30, 2026

Alethia: A Foundational Encoder for Voice Deepfakes

Existing voice deepfake detection and localization models rely heavily on representations extracted from speech foundation models (SFMs). However, downstream finetuning has now reached a state of diminishing returns. In this paper, we shift the focus to pretraining and propose a novel recipe that combines bottleneck masked embedding prediction with flow-matching based spectrogram reconstruction. The outcome, Alethia, is the first foundational audio encoder for various voice deepfake detection and localization tasks. We evaluate on

5

different tasks with

56

benchmark datasets, and note Alethia significantly outperforms state-of-the-art SFMs with superior robustness to real-world perturbations and zero-shot generalization to unseen domains (e.g., singing deepfakes). We also demonstrate the limitation of discrete targets in masked token prediction, and show the importance of continuous embedding prediction and generative pretraining for capturing deepfake artifacts.

Yi Zhu, Brahmi Dwivedi, Jayaram Raghuram +1

Reality Defender Inc.

Audio Representation Learning Audio Deepfake Detection Speech Foundation Models Audio Deepfake Detection Robustness

Apr 30, 2026

From Birdsong to Rumbles: Classifying Elephant Calls with Out-of-Species Embeddings

We show that pretrained acoustic embeddings classify elephant vocalisations at a level approaching that of end-to-end supervised neural networks, without any fine-tuning of the embedding model. This result is of practical importance because annotated bioacoustic data are scarce and costly to obtain, leaving conventional supervised approaches prone to overfitting and to poor generalisation under domain shift. A broad range of embedding models drawn from general audio, speech, and bioacoustic domains is evaluated, all of which are either out-of-domain (containing no bioacoustic data) or out-of-species (containing no elephant call data). The embedding networks themselves remain fixed; only the lightweight downstream classifiers, which include a linear model and several small neural networks, are trained. Among the models considered, Perch 2.0 achieves the best cross-validated classification performance, attaining AUCs of 0.849 on African bush elephant (Loxodonta africana) calls and 0.936 on Asian elephant (Elephas maximus) calls, with Perch 1.0 close behind. The best-performing system is within 2.2 % of an end-to-end supervised elephant call classification system. A layerwise analysis of pretrained transformer encoders, considered as embedding models, shows that intermediate representations outperform final-layer outputs. The second layer of both wav2vec2.0 and HuBERT encodes sufficient information for effective elephant call classification; truncation at this layer therefore preserves classification performance whilst retaining only approximately 10 % of the parameters of the full network. Such compact embedding networks are well suited to on-device processing where computational resources are limited.

Christiaan M. Geldenhuys, Thomas R. Niesler

Department of Electrical and Electronic Engineering University of Stellenbosch

Audio Representation Learning Audio Understanding Bioacoustics Audio Classification

Apr 30, 2026

Beyond the Baseband: Adaptive Multi-Band Encoding for Full-Spectrum Bioacoustics Classification

Animals hear and vocalize across frequency ranges that differ substantially from humans, often extending into the ultrasonic domain. Yet most computational bioacoustics systems rely on audio models pre-trained at 16 kHz, restricting their usable bandwidth to the 0-8 kHz baseband and discarding higher-frequency information present in many bioacoustic recordings. We investigate a multi-band encoding framework that decomposes the full spectrum of animal calls into band features and fuses them into a unified representation. Similarity analyses on models show that certain encoders produce decorrelated band embeddings that improve class separation after fusion. Classification experiments on three bioacoustic datasets using eight pre-trained models and five fusion strategies show that fused representations consistently outperform the baseband and time-expansion baselines on two datasets, showing the potential of multi-band methods for full-spectrum encoding of animal calls.

Eklavya Sarkar, Marius Miron, David Robinson +6

Earth Species Project

Audio Representation Learning Frequency-Domain Feature Learning Bioacoustics Audio Classification

Apr 29, 2026

Multimodal LLMs are not all you need for Pediatric Speech Language Pathology

Speech Sound Disorders (SSD) affect roughly five percent of children, yet speech-language pathologists face severe staffing shortages and unmanageable caseloads. We test a hierarchical approach to SSD classification on the granular multi-task SLPHelmUltraSuitePlus benchmark. We propose a cascading approach from binary classification to type, and symptom classification. By fine-tuning Speech Representation Models (SRM), and using targeted data augmentation we mitigate biases found by previous works, and improve upon all clinical tasks in the benchmark. We also treat Automatic Speech Recognition (ASR) with our data augmentation approach. Our results demonstrate that SRM consistently outperform the LLM-based state-of-the-art across all evaluated tasks by a large margin. We publish our models and code to foster future research.

Darren Fürst, Sebastian Steindl, Ulrich Schäfer

Ostbayerische Technische Hochschule Amberg-Weiden

Audio Representation Learning Healthcare Automatic Speech Recognition

Apr 27, 2026

S-SONDO: Self-Supervised Knowledge Distillation for General Audio Foundation Models

General audio foundation models have recently achieved remarkable progress, enabling strong performance across diverse tasks. However, state-of-the-art models remain extremely large, often with hundreds of millions of parameters, leading to high inference costs and limited deployability on edge devices. Knowledge distillation is a proven strategy for model compression, but prior work in audio has mostly focused on supervised settings, relying on class logits, intermediate features, or architecture-specific techniques. Such assumptions exclude models that output only embeddings, such as self-supervised or metric-learning models. We introduce S-SONDO (Self-Supervised KnOwledge DistillatioN for General AuDio FOundation Models), the first framework to distill general audio models using only their output embeddings. By avoiding the need for logits or layer-level alignment, S-SONDO is architecture-agnostic and broadly applicable to embedding-based teachers. We demonstrate its effectiveness by distilling two audio foundation models into three efficient students that are up to 61 times smaller while retaining up to 96% of teacher performance. We also provide practical insights on loss choice and clustering-based balanced data sampling. Code is available here: https://github.com/MedAliAdlouni/ssondo.

Mohammed Ali El Adlouni, Aurian Quelennec, Pierre Chouteau +2

LTCI, T´el´ecom Paris, Institut Polytechnique de Paris, Palaiseau, France

Model Compression Audio Representation Learning Self-Supervised Knowledge Distillation Knowledge Distillation

Apr 22, 2026

Enhancing Speaker Verification with Whispered Speech via Post-Processing

Speaker verification is a task of confirming an individual's identity through the analysis of their voice. Whispered speech differs from phonated speech in acoustic characteristics, which degrades the performance of speaker verification systems in real-life scenarios, including avoiding fully phonated speech to protect privacy, disrupt others, or when the lack of full vocalization is dictated by a disease. In this paper we propose a model with a training recipe to obtain more robust representations against whispered speech hindrances. The proposed system employs an encoder--decoder structure built atop a fine-tuned speaker verification backbone, optimized jointly using cosine similarity--based classification and triplet loss. We gain relative improvement of 22.26% compared to the baseline (baseline 6.77% vs ours 5.27%) in normal vs whispered speech trials, achieving AUC of 98.16%. In tests comparing whispered to whispered, our model attains an EER of 1.88% with AUC equal to 99.73%, which represents a 15% relative enhancement over the prior leading ReDimNet-B2. We also offer a summary of the most popular and state-of-the-art speaker verification models in terms of their performance with whispered speech. Additionally, we evaluate how these models perform under noisy audios, obtaining that generally the same relative level of noise degrades the performance of speaker verification more significantly on whispered speech than on normal speech.

Magdalena Gołębiowska, Piotr Syga

Department of Artificial Intelligence, Wroclaw University of Science and Technology, Wybrzeze Wyspianskiego 27, Wroclaw, 50-370, Poland

Audio Representation Learning Speaker Verification Speech Processing

Apr 21, 2026

Audio Spoof Detection with GaborNet

An direction of development in the extraction of features from audio signals is based on processing raw samples in the time domain. Such an approach appears to be effective, especially in the era of neural networks. An example is SincNet. In this solution, the core of the neural network layer is a set of sinc functions that are convolved with the input signal. Due to the finite length of sinc functions, distortions appear in the frequency domain of the convolved signal, the same as in the case of windowing the signal. Recently, a new approach has been developed that uses Gabor filters to replace sinc functions. Due to the complex results, further modifications had to be applied, such as squared modulus or Gaussian Lowpass Pooling. In this work, an ingestion layer based on a bank of Gabor filters, named GaborNet, and its modifications are intensively examined within the popular RawNet2 and RawGAT- ST architectures. These have been developed for the purpose of audio spoof detection. Another issue that has been investigated was audio augmentation using codec conversions, room responses, and additive noises.

Waldek Maciejko

Audio Representation Learning Audio Deepfake Detection

Apr 14, 2026

On the Distillation Loss Functions of Speech VAE for Unified Reconstruction, Understanding, and Generation

Continuous speech representations based on Variational Autoencoders (VAEs) have emerged as a promising alternative to traditional spectrogram or discrete token based features for speech generation and reconstruction. Recent research has tried to enrich the structural information in VAE latent representations by aligning with self-supervised learning (SSL) features, aiming for better generation performance. However, it remains unclear whether the widely-used alignment approach based on time-axis distillation is optimal when considering more tasks. To address this problem, this paper systematically explores different alignment approaches and analyzes their impact on the performances over three axes: reconstruction, understanding, and generation. We investigate various design choices in the distillation loss. Extensive experiments show that the joint-marginal alignment approach with adaptive weighting can achieve the best overall performance while allowing for a controllable balance.

Changhao Cheng, Wei Wang, Wangyou Zhang +4

School of Artificial Intelligence, Shanghai Jiao Tong University, Shanghai, China · Auditory Cognition and Computational Acoustics Lab, MoE Key Lab of Artificial Intelligence, AI Institute, School of Computer Science, Shanghai Jiao Tong University, Shanghai, China · ByteDance Seed, China

Variational Autoencoders Audio Representation Learning Knowledge Distillation Self-Supervised Speech Representation Learning

Apr 12, 2026

Cross-Cultural Bias in Mel-Scale Representations: Evidence and Alternatives from Speech and Music

Modern audio systems universally employ mel-scale representations derived from 1940s Western psychoacoustic studies, potentially encoding cultural biases that create systematic performance disparities. We present a comprehensive evaluation of cross-cultural bias in audio front-ends, comparing mel-scale features with learnable alternatives (LEAF, SincNet) and psychoacoustic variants (ERB, Bark, CQT) across speech recognition (11 languages), music analysis (6 collections), and European acoustic scene classification (10 European cities). Our controlled experiments isolate front-end contributions while holding architecture and training protocols minimal and constant. Results demonstrate that mel-scale features yield 31.2% WER for tonal languages compared to 18.7% for non-tonal languages (12.5% gap), and show 15.7% F1 degradation between Western and non-Western music. Alternative representations significantly reduce these disparities: LEAF reduces the speech gap by 34% through adaptive frequency allocation, CQT achieves 52% reduction in music performance gaps, and ERB-scale filtering cuts disparities by 31% with only 1% computational overhead. We also release FairAudioBench, enabling cross-cultural evaluation, and demonstrate that adaptive frequency decomposition offers practical paths toward equitable audio processing. These findings reveal how foundational signal processing choices propagate bias, providing crucial guidance for developing inclusive audio systems.

Shivam Chauhan, Ajay Pundhir

Presight AI, Abu Dhabi, United Arab Emirates

Audio Representation Learning Audio Classification ASR Evaluation

Apr 12, 2026

Whisper-AuT: Domain-Adapted Audio Encoder for Efficient Audio-LLM Training

Audio-native large language models (audio-LLMs) commonly use Whisper as their audio encoder. However, Whisper was trained exclusively on speech data, producing weak representations for music and environmental sound. This forces downstream audio-LLMs to compensate through extensive training on large-scale non-speech data. We present Whisper-AuT, a domain-adapted audio encoder obtained by fine-tuning Whisper-large-v3 on a curated mixture of speech (80%), environmental sound (10%), and music (10%) totaling approximately 20M samples. The full encoder-decoder is trained end-to-end with a seq2seq captioning objective; the decoder is then discarded and only the encoder is retained. Linear probe evaluations show that Whisper-AuT achieves +23.0% on ESC-50 (environmental sound), +5.0% on GTZAN (music genre), and +0.7% on Speech Commands (keyword spotting) compared to the original Whisperlarge-v3 encoder. Whisper-AuT is designed as a drop-in replacement for Whisper in audio-LLM architectures, with the goal of reducing downstream training cost by providing stronger initial audio representations for non-speech domains.

Jielin Qiu, Ming Zhu, Wenting Zhao +11

Salesforce AI Research

Audio Representation Learning Domain Adaptation Audio-Language Models

Apr 11, 2026

Masked Contrastive Pre-Training Improves Music Audio Key Detection

Self-supervised music foundation models underperform on key detection, which requires pitch-sensitive representations. In this work, we present the first systematic study showing that the design of self-supervised pretraining directly impacts pitch sensitivity, and demonstrate that masked contrastive embeddings uniquely enable state-of-the-art (SOTA) performance in key detection in the supervised setting. First, we discover that linear evaluation after masking-based contrastive pretraining on Mel spectrograms leads to competitive performance on music key detection out of the box. This leads us to train shallow but wide multi-layer perceptrons (MLPs) on features extracted from our base model, leading to SOTA performance without the need for sophisticated data augmentation policies. We further analyze robustness and show empirically that the learned representations naturally encode common augmentations. Our study establishes self-supervised pretraining as an effective approach for pitch-sensitive MIR tasks and provides insights for designing and probing music foundation models.

Ori Yonay, Tracy Hammond, Tianbao Yang

Department of Computer Science and Engineering, Texas A&M University

Music Information Retrieval Audio Representation Learning Self-Supervised Pre-Training Contrastive Learning

Apr 3, 2026

Unmixing The Crowd: Learning Persistent Speaker Representations from Mixture-Derived Multi-Speaker Embeddings

We study whether persistent conversational speaker structure can be extracted directly from local overlapping speech mixtures. We propose a teacher-student framework that learns mixture-derived multi-speaker embeddings using only short overlapping segments and permutation-invariant latent supervision. Despite never being explicitly trained for speaker tracking, diarization, or conversational memory, the learned embedding space supports long-form speaker re-identification when combined with a lightweight online memory mechanism during inference. We additionally observe that the learned representation retains meaningful speaker structure under unseen overlap cardinalities. We further show that embeddings extracted from separation-first pipelines exhibit degraded clustering structure compared to embeddings predicted directly from mixtures. Finally, the learned embeddings remain effective for the downstream target speaker extraction task across multiple architectures. These findings suggest that local mixture-derived representations support persistent conversational speaker re-identification when combined with lightweight inference-time memory consolidation.

Sidharth Sidharth, Meysam Asgari, Hao-Wen Dong +1

University of Michigan, Ann Arbor, MI, USA · Oregon Health & Science University, Portland, OR, USA

Audio Representation Learning Biometric Identification

Apr 1, 2026

FineLAP: Taming Heterogeneous Supervision for Fine-grained Language-Audio Pretraining

Contrastively pretrained audio-language models (e.g., CLAP) excel at clip-level understanding but struggle with frame-level tasks. Existing extensions fail to exploit the varying granularity of real-world audio-text data, where massive clip-level textual descriptions coexist with limited frame-level annotations. This paper proposes Fine-grained Language-Audio Pretraining (FineLAP), a novel training paradigm that advances both clip- and frame-level alignment in CLAP with heterogeneous data. FineLAP introduces a dual-stream sigmoid loss with a cluster-based sampling strategy to jointly learn from clip- and frame-level supervision. To capture both global semantics and local details, FineLAP uses a decoupled audio projector on top of a self-supervised encoder. To alleviate the scarcity of temporally annotated data, we present FineLAP-100k, a large-scale synthetic SED dataset constructed through a scalable curation pipeline. Extensive experiments demonstrate that FineLAP achieves SOTA performance across multiple audio understanding tasks, including retrieval, classification, sound event detection, and text-to-audio grounding. Ablation studies further show that coarse- and fine-grained alignment are mutually beneficial, providing insights for building better audio-language models (ALMs).

Xiquan Li, Xuenan Xu, Ziyang Ma +4

X-LANCE Lab, Shanghai Jiao Tong University, China · SJTU Paris Elite Institute of Technology, Shanghai Jiao Tong University, China · Shanghai Innovation Institute, China +1

Audio-Language Models Audio Representation Learning Temporal Audio Grounding Sound Event Detection

Mar 31, 2026

Combining Masked Language Modeling and Cross-Modal Contrastive Learning for Prosody-Aware TTS

Text encoders in neural text-to-speech (TTS) see only linguistic content, so prosody must be inferred from paired data and synthesised intonation is often flat. Contrastive text-audio pretraining (CLAPSpeech) injects prosodic knowledge by contrasting acoustic realisations of the same phoneme, but it never trains the encoder to tell different phonemes apart, and its interaction with masked language modelling (MLM) is unexplored. We study a curriculum for a speaker-conditioned phoneme-and-subword encoder: MLM, contrastive alignment over batches of different phonemes, and CLAPSpeech-style same-phoneme refinement. On Russian conversational speech, with each variant conditioning an identical Grad-TTS model, MLM followed by mixed-phoneme alignment gives the lowest word error rate (17.6% versus 32.8% for same-phoneme training alone) and the best quality and intonation ratings. Adding same-phoneme refinement improves same-phoneme retrieval yet yields the worst synthesis: embedding-space retrieval does not predict usefulness for generation.

Kirill Borodin, Vasiliy Kudryavtsev, Maxim Maslov +3

MTUCI, Moscow, Russia

TTS Synthesis Speech Prosody Audio Representation Learning Audio Diffusion Models Multimodal Contrastive Learning Masked Language Modeling

Mar 28, 2026

Can pre-trained Deep Learning models predict groove ratings?

This study explores the extent to which deep learning models can predict groove and its related perceptual dimensions directly from audio signals. We critically examine the effectiveness of seven state-of-the-art deep learning models in predicting groove ratings and responses to groove-related queries through the extraction of audio embeddings. Additionally, we compare these predictions with traditional handcrafted audio features. To better understand the underlying mechanics, we extend this methodology to analyze predictions based on source-separated instruments, thereby isolating the contributions of individual musical elements. Our analysis reveals a clear separation of groove characteristics driven by the underlying musical style of the tracks (funk, pop, and rock). These findings indicate that deep audio representations can successfully encode complex, style-dependent groove components that traditional features often miss. Ultimately, this work highlights the capacity of advanced deep learning models to capture the multifaceted concept of groove, demonstrating the strong potential of representation learning to advance predictive Music Information Retrieval methodologies.

Axel Marmoret, Nicolas Farrugia, Jan Alexander Stupacher

IMT Atlantique, Lab-STICC, UMR CNRS 6285, Brest, France · Center for Music In the Brain, Aarhus University, Århus, Denmark

Music Information Retrieval Audio Representation Learning

Mar 28, 2026

Unsupervised Evaluation of Deep Audio Embeddings for Music Structure Analysis

Music Structure Analysis (MSA) aims to uncover the high-level organization of musical pieces. State-of-the-art methods are often based on supervised deep learning, but these methods are bottlenecked by the need for heavily annotated data and inherent structural ambiguities. In this paper, we propose an unsupervised evaluation of nine open-source, generic pre-trained deep audio models, on MSA. For each model, we extract barwise embeddings and segment them using three unsupervised segmentation algorithms (Foote's checkerboard kernels, spectral clustering, and Correlation Block-Matching (CBM)), focusing exclusively on boundary retrieval. Our results demonstrate that modern, generic deep embeddings generally outperform traditional spectrogram-based baselines, but not systematically. Furthermore, our unsupervised boundary estimation methodology generally yields stronger performance than recent linear probing baselines. Among the evaluated techniques, the CBM algorithm consistently emerges as the most effective downstream segmentation method. Finally, we highlight the artificial inflation of standard evaluation metrics and advocate for the systematic adoption of trimming'', or even double trimming'' annotations to establish more rigorous MSA evaluation standards.

Axel Marmoret

BRAIN team, MEE, IMT Atlantique, Brest, France

Audio Representation Learning Music Information Retrieval

Mar 27, 2026

A Human-Inspired Decoupled Architecture for Efficient Audio Representation Learning

While self-supervised learning (SSL) has revolutionized audio representation, the excessive parameterization and quadratic computational cost of standard Transformers limit their deployment on resource-constrained devices. To address this bottleneck, we propose HEAR (Human-inspired Efficient Audio Representation), a novel decoupled architecture. Inspired by the human cognitive ability to isolate local acoustic features from global context, HEAR splits the processing pipeline into two dedicated modules: an Acoustic Model for local feature extraction and a Task Model for global semantic integration. Coupled with an Acoustic Tokenizer trained via knowledge distillation, our approach enables robust Masked Audio Modeling (MAM). Extensive experiments demonstrate that HEAR requires only 15M parameters and 9.47 GFLOPs for inference, operating at a fraction of the computational cost of conventional foundation models (which typically require 85M-94M parameters). Despite this high efficiency, HEAR achieves highly competitive performance across diverse audio classification benchmarks. The code and pre-trained models are available at https://github.com/HarunoriKawano/HEAR

Harunori Kawano, Takeshi Sasaki

Faculty of Engineering and Information Technology, University of Technology Sydney · Dept. of Electrical Engineering and Computer Science, Shibaura Institute of Technology

Audio Representation Learning Self-Supervised Pre-Training

Mar 26, 2026

Unlocking Strong Supervision: A Data-Centric Study of General-Purpose Audio Pre-Training Methods

Current audio pre-training seeks to learn unified representations for broad audio understanding tasks, but it remains fragmented and is fundamentally bottlenecked by its reliance on weak, noisy, and scale-limited labels. Drawing lessons from vision's foundational pre-training blueprint, we argue that the audio field must first establish its own large-scale, strong supervision framework. We introduce a new data-centric pipeline that leverages a high-fidelity captioner to create SOTA-quality captions and the first Unified Tag System (UTS) that bridges speech, music, and environmental sounds. We then conduct a systematic comparative study of different pre-training objectives on these strong source data. Our experiments suggest that data quality and coverage are the primary drivers of performance, while the choice of objective dictates downstream task specialization.

Xuanru Zhou, Yiwen Shao, Wei-Cheng Tseng +1

UT-Austin, Austin, USA · Tencent AI Lab, Bellevue, USA

Audio Representation Learning Audio Classification

Mar 25, 2026

Rethinking Masking Strategies for Masked Prediction-based Audio Self-supervised Learning

Since the introduction of Masked Autoencoders, various improvements to masking techniques have been explored. In this paper, we rethink masking strategies for audio representation learning using masked prediction-based self-supervised learning (SSL) on general audio spectrograms. While recent informed masking techniques have attracted attention, we observe that they incur substantial computational overhead. Motivated by this observation, we propose dispersion-weighted masking (DWM), a lightweight masking strategy that leverages the spectral sparsity inherent in the frequency structure of audio content. Our experiments show that inverse block masking, commonly used in recent SSL frameworks, improves audio event understanding performance while introducing a trade-off in generalization. The proposed DWM alleviates these limitations and computational complexity, leading to consistent performance improvements. This work provides practical guidance on masking strategy design for masked prediction-based audio representation learning.

Daisuke Niizumi, Daiki Takeuchi, Masahiro Yasuda +3

Tokyo Metropolitan University, Tokyo, Japan · NTT Inc., Atsugi, Japan

Audio Representation Learning Self-Supervised Pre-Training Masked Autoencoders Audio Understanding

Mar 24, 2026

The Interspeech 2026 Audio Encoder Capability Challenge for Large Audio Language Models

This paper presents the Interspeech 2026 Audio Encoder Capability Challenge, a benchmark specifically designed to evaluate and advance the performance of pre-trained audio encoders as front-end modules for Large Audio Language Models (LALMs). While LALMs have shown remarkable understanding of complex acoustic scenes, their performance depends on the semantic richness of the underlying audio encoder representations. This challenge addresses the integration gap by providing a unified generative evaluation framework, XARES-LLM, which assesses submitted encoders across a diverse suite of downstream classification and generation tasks. By decoupling encoder development from LLM fine-tuning, the challenge establishes a standardized protocol for general-purpose audio representations that can effectively be used for the next generation of multimodal language models.

Heinrich Dinkel, Jiahao Zhou, Guanbo Wang +8

MiLM Plus, Xiaomi Inc., Beijing, China · DataoceanAI Inc., USA · Centre for Vision Speech and Signal Processing, University of Surrey, Guildford, UK +1

Audio-Language Model Evaluation Audio Representation Learning Audio-Language Models

Mar 23, 2026

AnimalCLAP: Taxonomy-Aware Language-Audio Pretraining for Species Recognition and Trait Inference

Animal vocalizations provide crucial insights for wildlife assessment, particularly in complex environments such as forests, aiding species identification and ecological monitoring. Recent advances in deep learning have enabled automatic species classification from their vocalizations. However, classifying species unseen during training remains challenging. To address this limitation, we introduce AnimalCLAP, a taxonomy-aware language-audio framework comprising a new dataset and model that incorporate hierarchical biological information. Specifically, our vocalization dataset consists of 4,225 hours of recordings covering 6,823 species, annotated with 22 ecological traits. The AnimalCLAP model is trained on this dataset to align audio and textual representations using taxonomic structures, improving the recognition of unseen species. We demonstrate that our proposed model effectively infers ecological and biological attributes of species directly from their vocalizations, achieving superior performance compared to CLAP. Our dataset, code, and models will be publicly available at https://dahlian00.github.io/AnimalCLAP_Page/.

Risa Shinoda, Kaede Shiohara, Nakamasa Inoue +2

The University of Osaka · The University of Tokyo · Institute of Science Tokyo

Audio-Language Models Bioacoustics Audio Representation Learning

Mar 22, 2026

HELIX: Scaling Raw Audio Understanding with Hybrid Mamba-Attention Beyond the Quadratic Limit

Audio representation learning typically evaluates design choices such as input frontend, sequence backbone, and sequence length in isolation. We show that these axes are coupled, and conclusions from one setting often do not transfer to others. We introduce HELIX, a controlled framework comparing pure Mamba, pure attention, and a minimal hybrid with a single attention bottleneck. All models are parameter-matched at about 8.3M parameters to isolate architectural effects. Across six datasets, we find that the preferred input representation depends on the backbone, and that attention hurts performance on short, stationary audio but becomes important at longer sequence lengths. On a 5-minute speaker identification task with 30,000 tokens, pure attention fails with out-of-memory errors, while HELIX closes an 11.5-point gap over pure Mamba.

Khushiyant, Param Thakkar

Department of Computer Science, University of Freiburg, Freiburg im Breisgau, Germany · Department of Computer Engineering and Information Technology, Veermata Jijabai Technological Institute, Mumbai, India

Audio Representation Learning Hybrid Mamba-Transformer Architectures Hybrid Attention Mamba

Mar 17, 2026

Evaluating Latent Space Structure in Timbre VAEs: A Comparative Study of Unsupervised, Descriptor-Conditioned, and Perceptual Feature-Conditioned Models

We present a comparative evaluation of latent space organization in three Variational Autoencoders (VAEs) for musical timbre generation: an unsupervised VAE, a descriptor-conditioned VAE, and a VAE conditioned on continuous perceptual features from the AudioCommons timbral models. Using a curated dataset of electric guitar sounds labeled with 19 semantic descriptors across four intensity levels, we assess each model's latent structure with a suite of clustering and interpretability metrics. These include silhouette scores, timbre descriptor compactness, pitch-conditional separation, trajectory linearity, and cross-pitch consistency. Our findings show that conditioning on perceptual features yields a more compact, discriminative, and pitch-invariant latent space, outperforming both the unsupervised and discrete descriptor-conditioned models. This work highlights the limitations of one-hot semantic conditioning and provides methodological tools for evaluating timbre latent spaces, contributing to the development of more controllable and interpretable generative audio models.

Joseph Cameron, Alan Blackwell

Department of Computer Science & Technology, University of Cambridge, Cambridge, United Kingdom

Variational Autoencoders Audio Representation Learning Audio Generation Conditional Variational Autoencoder

Mar 16, 2026

Spectrogram features for audio and speech analysis

Spectrogram-based representations have grown to dominate the feature space for deep learning audio analysis systems, and are often adopted for speech analysis also. Initially, the primary motivator for spectrogram-based representations was their ability to present sound as a two dimensional signal in the time-frequency plane, which not only provides an interpretable physical basis for analysing sound, but also unlocks the use of a wide range of machine learning techniques such as convolutional neural networks, that had been developed for image processing. A spectrogram is a matrix characterised by the resolution and span of its two dimensions, as well as by the representation and scaling of each element. Many possibilities for these three characteristics have been explored by researchers across numerous application areas, with different settings showing affinity for various tasks. This paper reviews the use of spectrogram-based representations and surveys the state-of-the-art to question how front-end feature representation choice allies with back-end classifier architecture for different tasks.

Ian McLoughlin, Lam Pham, Yan Song +7

Singapore Institute of Technology · Austrian Institute of Technology · The University of Science and Technology of China +1

Audio Representation Learning Frequency-Domain Feature Learning Speech Processing

Mar 14, 2026

Probing neural audio codecs for distinctions among English nuclear tunes

State-of-the-art spoken dialogue models (Défossez et al. 2024; Schalkwyk et al. 2025) use neural audio codecs to "tokenize" audio signals into a lower-frequency stream of vectorial latent representations, each quantized using a hierarchy of vector codebooks. A transformer layer allows these representations to reflect some time- and context-dependent patterns. We train probes on labeled audio data from Cole et al. (2023) to test whether the pitch trajectories that characterize English phrase-final (nuclear) intonational tunes are among these patterns. Results: Linear probes trained on the unquantized latents or some of the associated codewords yield above-chance accuracy in distinguishing eight phonologically specified nuclear tunes with monotonal pitch accents (top average test accuracy (TATA): 0.31) and the five clusters of these tunes that are robust in human speech production and perception (TATA: 0.45). Greater accuracy (TATAs: 0.74-0.89) is attained for binary distinctions between classes of rising vs. falling tunes, respectively used for questions and assertions. Information about tunes is spread among all codebooks, which calls into question a distinction between 'semantic' and 'acoustic' codebooks found in the literature. Accuracies improve with nonlinear probes, but discrimination among the five clusters remains far from human performance, suggesting a fundamental limitation of current codecs.

Juan Pablo Vigneaux, Jennifer Cole

Northwestern University

Speech Prosody Neural Audio Codecs Audio Representation Learning

Mar 14, 2026

Multimodal Emotion Regression with Multi-Objective Optimization and VAD-Aware Audio Modeling for the 10th ABAW EMI Track

We participated in the 10th ABAW Challenge, focusing on the Emotional Mimicry Intensity (EMI) Estimation track on the Hume-Vidmimic2 dataset. This task aims to predict six continuous emotion dimensions: Admiration, Amusement, Determination, Empathic Pain, Excitement, and Joy. Through systematic multimodal exploration of pretrained high-level features, we found that, under our pretrained feature setting, direct feature concatenation outperformed the more complex fusion strategies we tested. This empirical finding motivated us to design a systematic approach built upon three core principles: (i) preserving modality-specific attributes through feature-level concatenation; (ii) improving training stability and metric alignment via multi-objective optimization; and (iii) enriching acoustic representations with a VAD-inspired latent prior. Our final framework integrates concatenation-based multimodal fusion, a shared six-dimensional regression head, multi-objective optimization with MSE, Pearson-correlation, and auxiliary branch supervision, EMA for parameter stabilization, and a VAD-inspired latent prior for the acoustic branch. On the official validation set, the proposed scheme achieved our best mean Pearson Correlation Coefficient of 0.478567.

Jiawen Huang, Chenxi Huang, Zhuofan Wen +7

School of Artificial Intelligence, University of Chinese Academy of Sciences · Institute of Automation, Chinese Academy of Sciences · Tianjin Normal University

Emotion Recognition Multimodal Emotion Recognition Multimodal Fusion Audio Representation Learning