STFT-based speech enhancement typically adopts overlapping analysis frames. While overlap is essential for stable STFT processing, it makes adjacent frames highly correlated, causing redundant computation in lightweight models. We propose Half-frame-rate Adaptive Learnable Operator (HALO), a causal plug-in module that halves the internal frame rate without altering the STFT procedure. Broadly applicable to many lightweight models, HALO applies adaptive rate reduction before the backbone and restoration afterward, reconstructing the full-rate spectrum on the original STFT grid. Both reduction and restoration are implemented with lightweight dynamic convolutions. By halving the processed frame rate, HALO reduces backbone compute cost with no added algorithmic latency, freeing budget for channel widening. Experiments on the DNS3 dataset show consistent gains across diverse lightweight models under matched complexity, demonstrating the effectiveness of reducing overlap-induced redundancy.

Spoken dialogue models typically start from text LLM backbones, yet reasoning often degrades when conditioning on speech instead of text. We attribute part of this modality gap to a temporal-granularity mismatch: speech tokens are temporally redundant and far longer than text under matched semantics, diluting per-token semantic density and weakening text-native reasoning dynamics. We study speech token design as a representation selection problem and sweep frame rates under a frozen LLM backbone with a fixed information rate. To make low frame rates feasible, we introduce factorized FSQ and a lightweight non-autoregressive audio LM head, scaling capacity to nearly 300\,bits/frame without sacrificing efficient prediction. With the bottleneck removed, we sweep frame rates (50$\rightarrow$2.08\,Hz) and alignment depth, and observe a consistent best regime for speech QA at 4.17\,Hz with intermediate-layer representation alignment.

This paper presents a novel data-free and training-free compression approach for speech foundation models using channelwise clustering via k-means. More fine-grained, mixed sparsity pruning by layer-level varying number of parameter clusters is also explored. Experiments conducted on the LibriSpeech dataset suggest that when operating with pruning sparsity of 50% on HuBERT-large, consistent WER reductions of 27.73%/18.61% absolute (34.37%/21.91% relative) over the magnitude-based pruning were obtained on the test-clean and test-other subsets before fine-tuning and 0.19%/0.79% absolute (3.36%/4.62% relative) after fine-tuning with only 3 epochs. Similar WER reductions of 2.86%/5.02% absolute (59.21%/55.29% relative) were observed against magnitudebased pruning on Whisper-large-v3 at 10% sparsity, all with no significant WER increase relative to the uncompressed baseline.

Multi-talker conversational automatic speech recognition data are often used to train speaker diarization models. Because such data prioritize semantic continuity, pauses and boundary margins are included within speech segments, resulting in loose annotations. Models trained on such data tend to internalize mechanisms that reproduce this looseness, although tight speech intervals are sometimes preferable for downstream applications. In this paper, we address the novel task of enabling models to produce tight predictions using loose labels. Our method generates tighter pseudo labels using causal and anticausal models, which are inherently incapable of learning loosening behavior. We further propose a co-training scheme that iteratively tightens labels and updates both models for more progressive refinement. Experimental results show that the proposed method recovers about 70 % of the tightening effect achieved by ideal tight-label training and improves downstream performance.

Distilling a large speech foundation model (SFM) into an efficient student model has been successfully applied to low-resource environments. Although distillation reduces inference latency, it requires an additional student model training. However, the training efficiency of SFM distillation remains underexplored. In this work, we explore training acceleration of SFM distillation to speed up model deployment. We examine the potential of stacking, in which the model depth is progressively increased through training until the target model depth is reached. While existing stacking methods improve training speed, they suffer from performance degradation. To handle this limitation, we propose interleaved stacking, a novel stacking method that consistently preserves layer position throughout the stacking process. This property is particularly critical in SFMs, in which each layer encodes distinct layer-specific knowledge. We validate the effectiveness of the proposed method on SUPERB.

Learning-based speech compression has achieved promising low-bitrate performance, but many neural speech codecs still describe quantized latents with preset-rate discrete symbols or apply entropy coding only after symbol generation. Such designs decouple representation learning from probability modeling, limiting their ability to exploit the non-uniform usage and temporal dependencies of learned speech latents. In this paper, we benchmark neural speech compression from a rate--distortion perspective and further investigate entropy-constrained coding for low-bitrate speech compression. We first formulate a unified learning-based speech coding pipeline and provide a benchmark-style analysis of recent neural speech codecs, showing that explicit probability modeling remains underexplored in learned speech compression. We then propose ECC, an Entropy-Constrained Codec that combines scalar quantization with a learned entropy model. ECC integrates hyperprior-based side information, channel-wise context modeling, latent residual prediction, and lightweight temporal modeling to estimate latent likelihoods for rate estimation during training and arithmetic coding during inference. To further improve low-bitrate efficiency, ECC introduces entropy skip, which omits highly predictable residual symbols using decoder-available scale estimates without transmitting additional skip masks. Extensive experiments show that ECC achieves a favorable low-bitrate rate--distortion trade-off over conventional and neural codec baselines, reducing BD-rate by 39.9% on ViSQOL and 76.3% on PESQ on average over two widely-used test sets. Ablation and diagnostic studies further validate the effectiveness of entropy modeling. Project Page: this https URL

Evaluating generative spatial audio for First-Order Ambisonics (FOA) remains challenging due to a limited understanding of how metrics respond to changes in spatial parameters such as azimuth and elevation. We propose a framework to analyze metric sensitivity along continuous spatial trajectories, drawing on principles of sensitivity analysis in parametric sound synthesis. Using controlled FOA scenes with increasing scene complexity, we define three desiderata for metric behavior: Responsiveness, Smoothness, and Symmetry. We assess standard distribution-based and sample-based metrics, including Fréchet Audio Distance (FAD), intensity vectors, and acoustic maps. Our findings show that FAD using localization-specific embeddings and acoustic maps yield high Responsiveness and robust Smoothness and Symmetry across conditions, while intensity vectors degrade with increasing scene complexity. This is the first step towards investigating the sensitivity of metrics for generative spatial audio.

Speech foundation models often struggle in low-resource domains due to domain mismatch and data scarcity. We propose Gumbel-BEARD, a domain adaptation framework that automates Whisper encoder layer selection via an end-to-end trainable hard Gumbel-Softmax selector. It enables self-supervised adaptation with a BEST-RQ objective that dynamically adapts to target acoustic characteristics without manual tuning. Experiments on the MyST child speech corpus demonstrate efficiency and scalability: with 10 h of labeled data for fine-tuning, our method matches a fully supervised baseline trained on the complete 133 h labeled set. We establish new state-of-the-art word error rates (WERs) of 8.21% using Whisper-medium on MyST and 11.06% using Whisper-small on the OGI Spontaneous dataset. Evaluation on CORAAL further confirms robustness to adult dialectal domain shifts, with up to 6% relative WER reduction, highlighting the generalizability of our approach to diverse low-resource conditions.

Large Audio-Language Models (LALMs) excel at audio understanding but expose little about where in an audio signal they attend. We introduce instruction-based vector steering, which constructs a steering vector by contrasting activations from differently instructed prompts while keeping the audio fixed. Through a systematic probe of LALM attention, we find that - unlike standard prompting or audio-based steering - this intervention significantly redistributes the temporal attention allocated to audio tokens, concentrating it on acoustically relevant regions. We then show that this attention shift is behaviorally meaningful: in a controlled three-event setting, reading out the temporal position of maximal steering-induced attention change recovers the location of a queried sound event without any training, attaining 60.87% and 68.72% overlap with ground-truth intervals on Qwen2-Audio and Audio Flamingo 3, far above direct prompting (31.84%, 46.75%) and random baselines (27.74%). Our results characterize a mechanistic property of instruction-based steering in LALMs and provide a training-free probe for the latent temporal structure these models encode.

Full-duplex spoken language models (FD-SLMs) enable seamless speech interaction by allowing models to listen and speak simultaneously, yet the internal mechanism by which they coordinate listening and speaking remains underexplored. We analyze the predictive behavior encoded in FD-SLM hidden representations and find that they exhibit stream-specific predictive patterns: during listening, they preferentially predict the incoming user stream, whereas during speaking, they preferentially predict the model output stream. Building on this observation, we show that FD-SLMs dynamically modulate their internal predictive focus between two states: a generative state aligned with model output generation and a perceptive state aligned with incoming user input. However, this modulation can lag behind abrupt changes in conversational context. During user interruptions, the model remains transiently biased toward the generative state before transitioning into the perceptive state, causing it to miss the beginning of the incoming input. We term this delayed internal transition state inertia. To quantify its downstream impact, we introduce the Zero-Buffer Benchmark (ZBB), a diagnostic benchmark for evaluating immediate interruption comprehension when user speech begins abruptly. We evaluate this setting using response correctness and initial-word occurrence rate (IWOR). Finally, we mitigate state inertia through activation steering with a perception vector, a training-free intervention with little additional computational overhead. Across multiple state-of-the-art FD-SLMs, activation steering substantially improves interruption handling; for example, on PersonaPlex, it improves correctness from 28% to 45% and IWOR from 40% to 72% without any fine-tuning.

Spoken language, whether produced by humans or large language models (LLM), unfolds over time with varying semantic content. However, we still lack simple, interpretable time-series features that capture how generic versus specific content is distributed over time, and that can be used to compare human and AI-generated speech. We introduce a semantic-timescale analysis pipeline that turns word-level transcripts with timestamps into semantic time-series. For each spoken narrative, we compute (i) semantic specificity using WordNet-based word depth and (ii) contextual similarity using SBERT embeddings and quantify their temporal dependence using autocorrelation-window measures (ACW-0 and related metrics). We then compare original speech to multiple shuffled controls that selectively disrupt lexical identity, temporal order, and word duration. Across human-read autobiographical narratives, TTS readings, and LLM-generated texts rendered with TTS, we find that segments with longer ACW-0 in the semantic time-series tend to contain more generic vocabulary, whereas segments with shorter ACW-0 are enriched in more specific words. These associations are strongly attenuated or abolished when word order and timing are randomized, indicating that ACW-based measures capture non-trivial temporal organization of semantic content beyond static lexical distributions. Our results suggest that ACW-based semantic timescales are a useful family of features for analyzing and comparing the temporal structure of human and AI-generated speech.

Automatic speech recognition systems have been shown to under-perform when it comes to transcribing words rarely seen in the training data, namely specialized terminology. Open-vocabulary keyword spotting, combined with contextual biasing, has been shown to mitigate this issue. However, existing systems can only handle glossaries of a few hundred terms without becoming an infeasible bottleneck. We propose a system that stores features with a memory footprint up to 128 times smaller than a comparable baseline and allows users to process massive databases while remaining open-vocabulary. Without fine-tuning the speech recognition model, our system achieves a comparable entity recall as uncompressed solutions, even in languages not seen during training.

Speech-based automatic estimation of depression levels is essential for enabling early detection and timely intervention, particularly in resource-constrained mental health settings. In recent years, deep learning has demonstrated impressive success across various domains, including affective computing and mental health assessment. Most existing approaches rely on RNN-based architectures (such as LSTM and GRU) to model temporal information for depression estimation. However, the extracted features often emphasize only a few adjacent speech segments, limiting their ability to capture long-range dependencies. To overcome this limitation, we introduce a memory-based feature augmentation method that enhances the representational capacity of GRU-extracted features. Rather than indiscriminately incorporating historical data, our memory bank is designed to selectively integrate two types of components in order to reduce redundancy and irrelevance: (1) historical temporal features that closely resemble the current GRU output, offering complementary contextual information; and (2) dynamic memory features identified based on feature variability, which capture behavioral and emotional fluctuations indicative of depressive symptoms. To effectively fuse the memory-augmented features with GRU outputs, we further design a Hierarchical Attention Fusion (HAF) module. Our method is evaluated on the widely used DAIC-WOZ and E-DAIC datasets, achieving state-of-the-art performance.

While Audio Language Models (ALMs) demonstrate strong semantic understanding, they struggle with complex affective interactions. Specifically, textual semantic dominance often overshadows acoustic nuances, and a lack of cognitive depth leads to generic, emotion-agnostic responses. We propose CogAudio-LLM\footnote{ \urlstyle{same} this https URL, a novel cognitive affective reasoning framework. To mitigate semantic dominance, we build LIME-440K, a ``lexically-identical, multi-emotion'' dataset designed to facilitate acoustic-semantic decoupling. We introduce EIPS, a 4-step Chain-of-Thought (CoT) mechanism incorporating psychological reasoning. For inference efficiency, multi-stage training explicitly establishes EIPS via supervised fine-tuning, then distills this logic into an implicit generation process. Finally, we design DR-SAPO (Dual-Route Soft Adaptive Policy Optimization) to dynamically balance the logical rigor of the CoT with the empathetic quality of the direct response.

Second-language (L2) speech recognition often requires transcriptions of pronunciations and intended meanings. Multi-task learning (MTL) is a natural approach because it assumes that shared representations benefit both outputs. However, this paper shows that this assumption does not hold across Korean and English. MTL improves meaning but degrades surface transcription, especially in English, where the degradation scales with surface-meaning divergence measured by Levenshtein edit distance. Encoder analysis links these patterns to encoder-level entanglement, with Korean preserving distinct task representations while English produces nearly identical ones. Cross-task decoder analysis shows that the meaning dual-output decoder adapts with a unique representation, while the surface dual-output decoder remains constrained by the encoder. These findings motivate the design of MTL frameworks that mitigate encoder-level entanglement to reduce surface degradation in dual-output L2 automatic speech recognition.

nnAudio is an open-source audio feature extraction toolbox for deep learning, but its use in current environments is hindered by TorchScript incompatibilities, inverse-transform edge cases, and dependency drift. We present a targeted modernization for modern PyTorch and scientific Python. We resolve TorchScript compilation failures in STFT and iSTFT by removing dynamic state mutation and module construction from scripted code paths and tightening argument handling in inverse-related helpers. We clarify inverse-STFT behavior by restricting reliable inversion to the uniform-bin setting (freq_scale=`no') and raising explicit runtime errors for unsupported frequency scales, preventing silently degraded reconstructions. We restore CFP compatibility with modern SciPy and ensure VQT reduces to CQT when gamma = 0. Regression tests cover the new STFT/iSTFT behaviors, and the updated codebase passes the full repository test suite in a modern Python environment. These improvements provide a more robust foundation for differentiable audio analysis in research and deployment.

Self-supervised speech representation learning has made significant progress through Siamese networks, which leverage different views of the same input. However, existing methods often require frame-wise alignment between these views, overlooking the broader linguistic context invariance across different speaking styles. We introduce SiamCTC, a framework that integrates Siamese networks with Connectionist Temporal Classification (CTC) to learn speech representations without strict frame-level correspondence. By employing CTC loss to establish flexible, monotonic alignments between differing temporal realizations of the same content, SiamCTC accommodates speed perturbations and other temporal augmentations. This design relaxes frame-wise constraints while preserving temporal coherence and enhancing robustness to speaking-rate variations in downstream tasks. Our experiments demonstrate that SiamCTC leads to more adaptable speech representations, particularly at diverse speaking rates.

We propose Relativistic Adversarial Feedback (RAF), a novel training objective for GAN vocoders that improves in-domain fidelity and generalization to unseen scenarios. Although modern GAN vocoders employ advanced architectures, their training objectives often fail to promote generalizable representations. RAF addresses this problem by leveraging speech self-supervised learning models to assist discriminators in evaluating sample quality, encouraging the generator to learn richer representations. Furthermore, we utilize relativistic pairing for real and fake waveforms to improve the modeling of the training data distribution. Experiments across multiple datasets show consistent gains in both objective and subjective metrics on GAN-based vocoders. Importantly, the RAF-trained BigVGAN-base outperforms the LSGAN-trained BigVGAN in perceptual quality using only 12\% of the parameters. Comparative studies further confirm the effectiveness of RAF as a training framework for GAN vocoders.

We present SAM, a State-space Audio-language Model that integrates an audio encoder with a Mamba-2 backbone. SAM-2.7B achieves 21.1 mAP on AudioSet and 17.6 SPICE on AudioCaps, matching or surpassing larger 7B transformer-based models with fewer parameters. We further provide the first systematic, representation-level analysis of how SSMs interact with audio encoder outputs: (1) joint audio encoder finetuning is essential, supported by accuracy gains and observed adaptation of token representation rank and similarity across different SSM sizes; (2) despite linear scaling, SSMs benefit more from compact, information-rich audio token representations than from excessively long token sequences; and (3) incorporating instruction-following supervision substantially improves reasoning ability, boosting MMAU-Sound accuracy from 22.8 to 56.8. Through comprehensive experiments and analysis, we establish practical design principles for SSMs as strong, scalable backbones for audio-language models.

Imperceptible text-based speech editing modifies spoken content through transcript manipulation while preserving acoustic continuity. Prior acoustic-space approaches suffer from content-style entanglement, causing unstable generation and boundary artifacts. We introduce a framework guided by the principle of "Edit Content, Preserve Acoustics". Editing is conducted in a stable semantic space, while acoustic realization is handled by a Flow Matching decoder. To ensure perceptual consistency, we propose Self-Consistency Rewards Group Relative Policy Optimization, which leverages a pre-trained Text-to-Speech model as an implicit critic, together with intelligibility and duration constraints. Experiments demonstrate consistent improvements over state-of-the-art autoregressive and non-autoregressive baselines in intelligibility, robustness, and perceptual quality.

We introduce LibriConvo, a synthetic conversational speech corpus for speaker diarization and automatic speech recognition (ASR), built by instantiating the previously proposed Speaker-Aware Simulated Conversation (SASC) framework in a dataset and benchmarking setting. The main contribution of this paper is a corpus construction pipeline and benchmark derived from that framework. To make the data more suitable for downstream ASR and diarization, conversational timing statistics are estimated from English CallHome using external voice activity detection, long pauses are compressed, LibriTTS utterances are grouped by book to improve local semantic continuity, and room impulse responses are selected with a spatial-plausibility heuristic. The resulting corpus contains 240.1 hours of audio across 1,496 dialogues involving 830 speakers, partitioned into speaker-disjoint train, validation, and test splits. We report baseline results for both diarization and ASR. On the test split, Sortformer outperforms the pyannote pipeline in diarization (11.1\% vs.~24.4\% DER). For ASR, a Fast Conformer-CTC XLarge model fine-tuned with Serialized Output Training achieves 7.29\% WER and 6.97\% cpWER, outperforming zero-shot Whisper-large-v3. These results position LibriConvo as a practical benchmark for studying synthetic conversational speech and for evaluating multi-speaker speech processing systems.

Automated deception detection is crucial for assisting humans in accurately assessing truthfulness and identifying deceptive behavior. Conventional contact-based techniques, like polygraph devices, rely on physiological signals to determine the authenticity of an individual's statements. Nevertheless, recent developments in automated deception detection have demonstrated that multimodal features derived from both audio and video modalities may outperform human observers on publicly available datasets. Despite these positive findings, the generalizability of existing audio-visual deception detection approaches across different scenarios remains largely unexplored. To close this gap, we present the first cross-domain audio-visual deception detection benchmark, that enables us to assess how well these methods generalize for use in real-world scenarios. We used widely adopted audio and visual features and different architectures for benchmarking, comparing single-to-single and multi-to-single domain generalization performance. To further exploit the impacts using data from multiple source domains for training, we investigate three types of domain sampling strategies, including domain-simultaneous, domain-alternating, and domain-by-domain for multi-to-single domain generalization evaluation. We also propose an algorithm to enhance the generalization performance by maximizing the gradient inner products between modality encoders, named ``MM-IDGM". Furthermore, we proposed the Attention-Mixer fusion method to improve performance, and we believe that this new cross-domain benchmark will facilitate future research in audio-visual deception detection.

Transformer encoder with connectionist temporal classification (CTC) framework is widely used for automatic speech recognition (ASR). However, knowledge distillation (KD) for ASR displays a problem of disagreement between teacher-student models in frame-level alignment which ultimately hinders it from improving the student model's performance. In order to resolve this problem, this paper introduces a self-knowledge distillation (SKD) method that guides the frame-level alignment during the training time. In contrast to the conventional method using separate teacher and student models, this study introduces a simple and effective method sharing encoder layers and applying the sub-model as the student model. Overall, our approach is effective in improving both the resource efficiency as well as performance. We also conducted an experimental analysis of the spike timings to illustrate that the proposed method improves performance by reducing the alignment disagreement.

Automatic speech recognition systems based on deep learning are mainly trained under empirical risk minimization (ERM). Since ERM utilizes the averaged performance on the data samples regardless of a group such as healthy or dysarthric speakers, ASR systems are unaware of the performance disparities across the groups. This results in biased ASR systems whose performance differences among groups are severe. In this study, we aim to improve the ASR system in terms of group robustness for dysarthric speakers. To achieve our goal, we present a novel approach, sample reweighting with sample affinity test (Re-SAT). Re-SAT systematically measures the debiasing helpfulness of the given data sample and then mitigates the bias by debiasing helpfulness-based sample reweighting. Experimental results demonstrate that Re-SAT contributes to improved ASR performance on dysarthric speech without performance degradation on healthy speech.