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

Silence is Golden: Mitigating Hallucinations in Large Audio-Language Models via Layer-Weighted Vector Steering

Large Audio-Language Models (LALMs) excel in Audio QA but often suffer from hallucinations ungrounded in the audio. To our knowledge, we are the first to propose applying vector steering to the audio domain to mitigate this. Unlike text-based steering, our silence-anchored contrastive approach steers the model away from hallucinations by contrasting active audio against a silent baseline. Probing internal states reveals a strong correlation between specific layer representations and output correctness. Leveraging this, we introduce Layer-Weighted Vector Steering (LWVS), a training-free intervention that increases steering strength at influential layers. On the Audio Hallucination QA dataset, LWVS significantly outperforms baselines, boosting Recall on the Gemma model by 15.6% (53.4% to 69.0%). Crucially, MMAU benchmark tests confirm LWVS preserves and even enhances general audio understanding, achieving an 8% relative accuracy increase on the Qwen model (54.8% to 59.2%).

cs.SD↗

arXiv2026

FastSLM: Hierarchical Temporal Abstraction for Efficient Long-Form Speech Adaptation

Scaling Multimodal Large Language Models (MLLMs) to long-form speech is bottlenecked by the explosive growth of input tokens. Existing speech-language models project high-frame-rate acoustic features directly into the LLM input space, making long-context processing computationally prohibitive. Unlike images or videos, speech lacks spatial redundancy, making extreme token compression particularly challenging. To address this limitation, we propose FastSLM, a token-efficient architecture featuring the Hierarchical Temporal Abstractor (HTA), which progressively distills acoustic features across multiple temporal scales. HTA achieves an extreme compression rate of 1.67 tokens per second (97% reduction) while preserving essential linguistic information for downstream speech-language understanding. Experimental results demonstrate that FastSLM achieves competitive performance across diverse speech-language tasks while requiring substantially fewer speech tokens and FLOPs than existing speech-language models. The source code and model checkpoints are available at https://github.com/Lee-junseok1025/FastSLM.

eess.AS↗

arXiv2026

MUSE: A Run-Centric Platform for Multimodal Unified Safety Evaluation of Large Language Models

Safety evaluation of multimodal large language models requires tracking not only whether an attack succeeds, but also how the interaction unfolds across turns and input modalities. We present MUSE (Multimodal Unified Safety Evaluation), an open-source, browser-based, run-centric platform for multimodal safety evaluation. MUSE treats each attack run as the persistent unit of execution, inspection, and analysis, preserving its configuration, multi-turn trajectory, delivered modalities and media, target responses, and safety judgments. A five-level response taxonomy further distinguishes full Compliance from Partial Compliance and refusal behavior, yielding hard ASR, soft ASR, and gray-zone width (GZW). Across 11,700 evaluations on six multimodal LLMs, direct text-only requests yield only 3.1% macro hard ASR and 4.4% soft ASR, while iterative attack procedures are substantially more effective. Attack effectiveness also varies substantially with the attacker backbone. In contrast, Inter-Turn Modality Switching (ITMS), evaluated as a controlled delivery-modality probe, does not consistently increase attack success. These results demonstrate the value of run-centric, fine-grained evaluation for characterizing multimodal safety behavior beyond a single binary success metric.

cs.LG↗

arXiv2026

KoALa-Bench: Evaluating Large Audio Language Models on Korean Speech Understanding and Faithfulness

Recent advances in large audio language models (LALMs) have enabled multilingual speech understanding. However, benchmarks for evaluating LALMs remain scarce for non-English languages, with Korean being one such underexplored case. In this paper, we introduce KoALa-Bench, a comprehensive benchmark for evaluating Korean speech understanding and speech faithfulness of LALMs. In particular, KoALa-Bench comprises six tasks. Four tasks evaluate fundamental speech understanding capabilities, including automatic speech recognition, speech translation, speech question answering, and speech instruction following, while the remaining two tasks evaluate speech faithfulness, motivated by our observation that several LALMs often fail to fully leverage the speech modality. Furthermore, to reflect Korea-specific knowledge, our benchmark incorporates listening questions from the Korean college scholastic ability test as well as content covering Korean cultural domains. We conduct extensive experiments across six models, including both white-box and black-box ones. Our benchmark, evaluation code, and leaderboard are publicly available at https://ksbench.github.io/Korean-Benchmark/.

cs.CL↗

arXiv2026

MRMAD: A Multi-Round Multi-Audio Benchmark for Evaluating Acoustic Degradation Perception in Large Audio-Language Models

Large audio-language models (LALMs) have shown promising progress in understanding speech, music, and general sound events, yet their ability to reason about how audio signals are degraded remains underexplored. Existing benchmarks primarily evaluate semantic understanding, event recognition, or high-level audio reasoning, leaving a basic question unanswered: Do LALMs understand the differences in audio quality? We introduce MRMAD, a Multi-Round Multi-Audio Degradation benchmark for evaluating audio degradation perception and understanding in LALMs. MRMAD spans speech, music, and sound, and frames evaluation as multi-turn dialogues across multiple audio inputs, requiring models to identify types of degradation, compare severity, and perceive corruption changes across turns. Unlike current single-turn audio-language benchmarks, MRMAD evaluates whether LALMs can maintain consistent degradation hypotheses with new evidence and comprehend low-level acoustic phenomena over multi-turn dialogues. Through a systematic evaluation of 18 representative LALMs from non-thinking to reasoning and Omni models, we find that current models often recognize coarse content while failing to diagnose, compare, or reason about degradations reliably. Human evaluations further reveal a significant perception gap between LALMs and human listeners. MRMAD thus exposes a critical yet overlooked aspect of audio-language understanding and provides a diagnostic foundation for building future LALMs that are robust to real-world acoustic conditions.

cs.SD↗

arXiv2026

Perceptually Better, Semantically Worse: Measuring Speech Enhancement Impact on LLM-Based Voice Systems

Speech enhancement (SE) is commonly applied as a preprocessing step in spoken AI pipelines under the assumption that better audio quality improves downstream task performance. Whether SE-induced distortions propagate to downstream LLM task performance remains an open question. We introduce Output Divergence Rate (ODR), which measures how often SE changes an LLM's intent classification relative to clean speech, and benchmark five conditions on 2,974 SLURP clips using Whisper large-v3 and wav2vec2-large cascades. Every condition produces ODR significantly above zero ($p < 0.001$, binomial test). MetricGAN{+} more than doubles ODR versus unenhanced noisy speech (0.318 vs. 0.135) despite improving PESQ, and unmitigated echo reaches an ODR of 0.836 through speaker substitution, a failure WER cannot capture. Audio quality metrics range from near-zero to moderate correlation with ODR (SQUIM-MOS $ρ=-0.068$, PESQ $ρ=-0.467$). The MetricGAN{+} and echo results replicate across ASR architectures, indicating that standard audio quality metrics are insufficient for LLM pipeline quality.

cs.SD↗

arXiv2026

Textual Acoustic Grounding for Generalizable LLM-Based Deepfake Voice Detection

Deepfake voice detection suffers from poor generalization across unseen domains. While Audio Large Language Models (ALLMs) show promise, the modality gap between continuous audio embeddings which capture the subtle acoustic details necessary for deepfake detection and the semantic space of LLMs remains a critical, underexplored bottleneck. We address this by benchmarking diverse audio encoders integrated with Qwen LLMs (0.5B to 7B parameters). First, we demonstrate that fine-tuning the LLM alone risks out-of-domain overfitting, making a frozen LLM a stronger, resource-efficient baseline. Second, to explicitly bridge the modality gap, we introduce a cross-modal prompting strategy that injects linguistic-knowledge-driven acoustic features (via openSMILE) as structured text tokens. This explicit textual grounding not only enhances the frozen baseline but also makes LLM fine-tuning more effective. Ultimately, our approach demonstrates state-of-the-art resilience on the out-of-domain ITW and MLAAD benchmarks, yielding over \textbf{16.2\%} absolute improvement in Macro-F1 over existing ALLM baselines while maintaining competitive in-domain performance. All models reported in this work are \href{https://huggingface.co/01Yassine/AudioLLM-Deepfake-Detection}{publicly available}.

cs.SD↗

arXiv2026

Conjoint Audio-to-Spikes Encoding and Processing for Efficient Neuromorphic Speech Recognition

Obtaining data from neuromorphic sensors and processing it with Spiking Neural Networks is a promising solution to lower the energy cost of artificial intelligence. The current rarity of natively neuromorphic datasets promotes the development of software tools to translate input sensory data into spikes. However, highly bio-mimetic simulators can be challenging to implement on digital hardware. In this work, we evaluate the neuromorphic encoding and subsequent classification of audio into spikes using a non-learnable, high-level, programmable encoder targeting hardware implementation on FPGA. We quantify the pipeline's efficiency with hardware-agnostic metrics based on the quantitative spiking activity. Our study focuses on the simultaneous optimisation of encoder and classifier: the first provides efficient and informative data so that the latter achieves a better performance with an overall lower energy cost at learning and inference. This work introduces the first end-to-end neuromorphic spike-encoding and evaluation of the TIMIT dataset. Our simple feedforward network reaches a classification accuracy of 99.77% on a spike-encoded Heidelberg Digits, overcoming the neuromorphic state of the art on this benchmark dataset.

cs.NE↗

arXiv2026

MusGU+: Toward a Musician-Centered Evaluation Framework and Discovery Tool for Generative Music AI

Generative music systems are increasingly presented as tools that democratize music creation, yet their practical suitability for musicians remains underexplored. Prior work includes openness-focused evaluation frameworks, such as MusGO (Music-Generative Open AI), as well as qualitative studies of musicians' experiences with generative systems. However, these approaches do not support systematic comparison or early-stage discovery of models for creative use. Motivated by such limitations, we introduce MusGU+, a musician-centered evaluation framework organized around three dimensions: Adaptability, Usability, and Controllability. Together, these capture whether a model can be feasibly trained or fine-tuned on personal data, integrated into real-world music workflows, and controlled in musically meaningful ways. We evaluate 10 representative generative music systems and present an interactive discovery tool that enables musicians to explore and filter models according to these criteria. While MusGO remains valuable for promoting responsible research practices, MusGU+ supports informed selection and practical adoption of generative systems by musicians.

cs.SD↗

arXiv2026

When Does Predictor-Based RL Align with Human Perception? A Study of Subjective Rewards in Codec-Based Speech Language Models

Codec-based text-to-speech (TTS) models make language-model post-training applicable to speech generation, but it remains unclear when learned perceptual predictors can serve as reinforcement learning rewards without losing alignment with human listeners. We study this question with Group Relative Policy Optimization (GRPO) using learned rewards for anime-like speaking style, naturalness, likability, and arousal. To prevent perceptual rewards from being optimized through transcript drift, we introduce a character error rate (CER) zone constraint and compare policy optimization with Best-of-$N$ reranking under the same reward gate. Across single-reward runs, each reward primarily improves its own target metric, showing that subjective predictors are not interchangeable quality surrogates. Multi-rater A/B tests further show uneven human transfer, while a reward-gap analysis separates average transfer from within-axis calibration: signed reward gaps significantly predict listener choices in the pooled analysis, whereas residual CER gaps do not, but per-axis calibration remains heterogeneous. Best-of-8 is a strong human-level baseline and is not clearly worse than GRPO perceptually, suggesting that GRPO should be viewed as amortizing reward-selected behavior into the policy rather than uniformly outperforming reranking. These results support analyzing subjective speech rewards as predictor-axis-base tuples and provide practical diagnostics for selecting rewards before multi-reward speech post-training.

cs.CL↗

arXiv2026

U-PAST: A Phase-Aware Audio Spectrogram Transformer-U-Net for Single-Channel Speech Enhancement

Convolutional neural networks (CNNs), used widely and successfully in audio enhancement, capture long-range time-frequency dependencies only indirectly, through successive convolution and pooling. Here, we present U-PAST, a hybrid transformer-U-Net architecture that addresses this limitation through self-attention dependency-modeling in the complex spectrogram domain. U-PAST tokenizes a complex STFT representation, similarly to the magnitude spectrogram tokenization of the Audio Spectrogram Transformer (AST), applies a multi-layer transformer encoder, and reconstructs the enhanced complex spectrogram with a U-Net-style decoder. We evaluate four architectural variants with between 1.17M and 2.40M parameters on the DNS Challenge, VoiceBank-DEMAND, and LibriMix corpora under matched, acoustic mismatch, and two-dataset mismatch conditions. U-PAST attains the best SI-SDR of any evaluated model under acoustic mismatch and closely trails substantially larger convolutional and time-domain baselines by 0.26 dB to 0.63 dB SI-SDR under the remaining three conditions while achieving the strongest perceptual (DNSMOS) quality under dataset mismatch. The largest evaluated configuration, U-PAST-H (2.40M parameters), is consistently the strongest variant of the family, offering an attractive performance-to-cost trade-off at a small parameter footprint.

eess.AS↗

arXiv2026

Annotation of Soft Onsets in String Ensemble Recordings

Onset detection is the process of identifying the start points of musical note events within an audio recording. While the detection of percussive onsets is often considered a solved problem, soft onsets-as found in string instrument recordings-still pose a significant challenge for state-of-the-art algorithms. The problem is further exacerbated by a paucity of data containing expert annotations and research related to best practices for curating soft onset annotations for string instruments. To this end, we investigate inter-annotator agreement between 24 participants, extend an algorithm for determining the most consistent annotator, and compare the performance of human annotators and state-of-the-art onset detection algorithms. Experimental results reveal a positive trend between musical experience and both inter-annotator agreement and performance in comparison with automated systems. Additionally, onsets produced by changes in fingering as well as those from the cello were found to be particularly challenging for both human annotators and automatic approaches. To promote research in best practices for annotation of soft onsets, we have made all experimental data associated with this study publicly available. In addition, we publish the ARME Virtuoso Strings dataset, consisting of over 144 recordings of professional performances of an excerpt from Haydn's string quartet Op. 74 No. 1 Finale, each with corresponding individual instrumental onset annotations.

eess.AS↗

arXiv2026

Joycent: Multi-Accent TTS via Disentangled Accent Modeling and Layer-Specific Conditioning

Accent text-to-speech (TTS) aims to synthesize speech with a target accent while preserving speaker identity, but faces two key challenges: disentangling accent from speaker characteristics and effectively conditioning speech generation on the two disentangled factors. In this paper, we propose Joycent, a diffusion-based accent TTS framework that addresses both challenges. Our key idea is to separate accent and speaker information in both representation learning and TTS conditioning. Joycent uses WhisAID, a Whisper-based accent encoder with gradient reversal to learn speaker-disentangled accent representations, and introduces layer-specific conditional layer normalization to inject accent and speaker information at different stages of the text encoder. We evaluate Joycent on the Mandarin Regional Accent Corpus (MRAC) with seen and unseen speakers, including a challenging cross-accent setting where the speaker and accent prompts come from different accents. Experimental results show that Joycent improves accent similarity over existing methods while maintaining strong speaker similarity, with consistent gains under the challenging cross-accent setting. Subjective evaluation further confirms improved naturalness, accent similarity, and speaker preservation. The audio samples are available at https://oshindow.github.io/joycent/.

cs.SD↗

arXiv2026

How Well Do Generative Music Models Follow Emotion Conditioning?

Recent generative music models offer increasingly fine-grained control through text and audio conditioning, yet how faithfully they follow intended emotional cues remains an open question. We address this gap with a unified evaluation pipeline for emotion-following in generated music. Using all 1000 tracks in GTZAN, we extract semantic audio descriptions with DashengLM, an audio captioning model, and estimate source-track valence and arousal with Music2Emotion, a music emotion recognition model. We construct affect-aware text prompts by combining descriptions with top-ranked emotion tags and generate 30-second outputs with three systems, Stable Audio Open, MusicGen, and InspireMusic, evaluating both text- and audio-conditioned generation. To measure emotion-following, we compute valence and arousal on generated audio and compare them with the source tracks using absolute error and Euclidean distance in valence-arousal space. Text-conditioned generation consistently outperforms audio conditioning, with MusicGen (text) and InspireMusic (text) achieving the best performance, while audio-conditioned variants prove less stable. We further find that valence is preserved more reliably than arousal and that emotion-following varies substantially across genres. These findings underscore the importance of evaluating affective controllability directly rather than relying solely on general quality or prompt-relevance metrics.

cs.SD↗

arXiv2026

UltraVoice: Scaling Fine-Grained Style-Controlled Speech Conversations for Spoken Dialogue Models

Spoken dialogue models currently lack the ability for fine-grained speech style control, a critical capability for human-like interaction that is often overlooked in favor of purely functional capabilities like reasoning and question answering. To address this limitation, we introduce UltraVoice, the first large-scale speech dialogue dataset engineered for multiple fine-grained speech style control. Encompassing over 830 hours of speech dialogues, UltraVoice provides instructions across six key speech stylistic dimensions: emotion, speed, volume, accent, language, and composite styles. Fine-tuning leading models such as SLAM-Omni and VocalNet on UltraVoice significantly enhances their fine-grained speech stylistic controllability without degrading core conversational abilities. Specifically, our fine-tuned models achieve improvements of 29.12-42.33% in Mean Opinion Score (MOS) and 14.61-40.09 percentage points in Instruction Following Rate (IFR) on multi-dimensional control tasks designed in the UltraVoice. Moreover, on the URO-Bench benchmark, our fine-tuned models demonstrate substantial gains in core understanding, reasoning, and conversational abilities, with average improvements of +10.84% on the Basic setting and +7.87% on the Pro setting. Furthermore, the dataset's utility extends to training controllable Text-to-Speech (TTS) models, underscoring its high quality and broad applicability for expressive speech synthesis. The complete dataset and model checkpoints are available at: https://github.com/bigai-nlco/UltraVoice.

eess.AS↗

arXiv2026

PairAlign: A Framework for Autoregressive Tokenization via Self-Alignment with Applications to Audio Tokenization

Modern learning systems represent perceptual signals with continuous vectors, but comparison, retrieval, memory, alignment, and reasoning are often symbolic. In language, tokens provide this interface; for speech and audio, it must be learned. Existing audio tokenizers rely on local quantization, clustering, or reconstruction, leaving sequence consistency, compactness, length, termination, and edit geometry only indirectly controlled. We introduce PairAlign, a framework for compact audio tokenization through autoregressive self-alignment. An encoder maps speech to a continuous condition, and an autoregressive decoder emits tokens from BOS to EOS. Given two content-preserving views, PairAlign derives a canonical anchor target and trains both views to predict it, with unrelated in-batch targets as competing sequences. It first learns an autoregressive bridge from VQ targets and then transitions to EMA-teacher self-alignment with grounding and anti-collapse controls. On 3 s speech, PairAlign produces shorter, non-degenerate sequences while retaining ordered paired-view consistency. On the TIMIT retrieval archive, it operates at 8.28 tokens/s, reducing archive token count by 70.65% and the edit-distance work proxy by 91.87% relative to Stage I. Positive-negative probes show strong separation across phone-disjoint, trigram-disjoint, and temporally rearranged negatives, while rate-controlled post-hoc BPE does not recover the same compactness-consistency operating point. These results expose a rate-granularity trade-off: PairAlign does not uniformly outperform denser tokenizers on every local metric, but provides a lower-rate symbolic interface preserving ordered and relational structure. Conceptually, PairAlign follows JEPA-style predictive learning by predicting an abstract target rather than reconstructing the input; here, the target is a learned variable-length symbolic sequence.

cs.LG↗

arXiv2026

FSA-GRPO: Teaching Auditory LLMs to Use Few-Shot Demonstrations

Few-shot prompting provides an effective way to adapt auditory large language models to low-resource tasks such as children's speech recognition. However, most auditory large language models are not explicitly trained to perform inference in this demonstration-conditioned format, limiting the extent to which they can benefit from In-Context Learning (ICL). To address this limitation, we introduce Few-Shot Aware GRPO (FSA-GRPO), an RL-based post-training recipe that uses a specially designed reward to encourage the model to leverage few-shot demonstrations, thereby strengthening its few-shot adaptation ability. Notably, training with only 2k high-resource adult ASR utterances improves the model's general few-shot adaptation ability, yielding gains not only in children's speech recognition (53.9% relative WER reduction without any in-domain training) but also in multilingual ASR (including low-resource languages), speech translation, and audio understanding. We further study data selection and the weight and similarity cutoffs of the auxiliary reward to identify an effective training recipe. Our experiments show that when in-domain data are unavailable or cannot be used for training, FSA-GRPO is more effective than direct tuning on related out-of-domain data.

eess.AS↗

arXiv2026

SpeechJBB: Probing Safety Alignment and Comprehension in Large Audio Language Models under Code-Switched Speech

Large audio language models (LALMs) are increasingly deployed in real-world applications, yet their safety alignment is still primarily evaluated on monolingual, text-based harmful prompts. This leaves their generalizability under multilingual and spoken settings, particularly code-switched speech, largely underexplored. To address this gap, we introduce SpeechJBB, an audio jailbreak dataset for benchmarking state-of-the-art LALMs across five European languages: English, French, German, Italian, and Spanish, as well as code-switched variants combining pairs of these languages. The extent of safety weaknesses is further probed by introducing an augmented setting where phonologically plausible pseudo-words are inserted around safety-critical terms to simulate localized obfuscation. Across models, code-switched harmful audio yields substantially high jailbreak success rates (JSR), with non-English monolingual and non-English code-switched pairs exhibiting the highest attack success. Pseudo-word insertion monotonically reduces refusal as insertion density increases, even though models rarely attribute harmful meaning to the inserted tokens. Comprehension benchmarks show that these failures are not reducible to multilingual misunderstanding, as several models with strong ASR, spoken language understanding, and spoken reasoning performance are among the most vulnerable.

cs.SD↗