ReasonAudio benchmark reveals that state-of-the-art text-audio retrieval models struggle with reasoning tasks like negation and duration, and multimodal LLMs lose reasoning ability after contrastive fine-tuning.
hub Mixed citations
author Zhou, A
Mixed citation behavior. Most common role is background (57%).
35
Pith papers citing it
925
external citations · external index
Background
57% of classified citations
hub tools
citation-role summary
background 4
method 2
dataset 1
citation-polarity summary
representative citing papers
ciwGAN and fiwGAN models trained on isolated words spontaneously generate concatenated multi-word outputs and display early compositionality precursors.
Sarashina2.2-TTS achieves SOTA kanji reading accuracy via data scaling and Joyo-kanji-targeted synthesis, introduces the Joyo Kanji Yomi Benchmark and Kana-CER metric, and shows stable cross-lingual performance.
Four self-stigma personas identified via LPA on 1,174 Reddit users; persona-conditioned LLMs achieve targeted shifts but experts prefer generic empathy baselines.
PolySpeech-100 is a new benchmark for native-level speech comprehension across 110 linguistic variants that evaluates 22 models and reports E2E advantages on dialects, robustness gaps on low-resource languages, and degradation from Chain-of-Thought prompting.
Predictive Entropy Maximization performs competitive blind source separation using only local error-driven and Hebbian updates derived from a surrogate entropy objective with spectral error bounds.
MARS is a transfer-based black-box attack that uses bi-level optimization on semantic and artifact anchors to escape the linearity trap and improve attack success rates on SSL-SVDD by up to 36%.
KL regularization aligning model predictions with empirical transition patterns improves macro-F1 by 9-42% in next dialogue act prediction on German counselling data and transfers to other datasets.
Introduces the LDD task, ListenForge dataset built from five listening head generation methods, and MANet model that detects listening forgeries via motion inconsistencies guided by audio semantics.
LLMs excel at retrospective mental-state labeling on naturalistic dialogues but mostly fail a context-free prospective probe that maps isolated mental-state profiles to dialogue trajectories, despite expert 100% accuracy.
CleanCodec reframes audio tokenization as a selective information bottleneck to encode only perceptually important features at 12.5 tokens per second, outperforming prior codecs in efficiency, speaker similarity, and intelligibility.
TB-AVA uses text-mediated gated semantic modulation to enable efficient audio-visual alignment, achieving state-of-the-art results on AVE, AVS, and AVVP benchmarks.
DECKER is a domain-invariant four-stage framework (keyboard normalization, adversarial disentanglement, cross-keyboard contrastive alignment, acoustic style randomization) plus LLM post-processing that improves keystroke inference over baselines on the new HEAR dataset, especially in cross-keyboard
SCALE disentangles emotion and cause representations in conversations and uses optimal transport for many-to-many global alignment, achieving SOTA on ECPEC benchmarks.
PRISM-CTG is the first large-scale foundation model for cardiotocography that uses multi-view self-supervised learning on unlabeled data to learn transferable representations, outperforming baselines on seven downstream tasks with external validation.
Momentum SGD pays a provable tracking penalty under distribution shift — lag growing as (1−β)⁻¹ and tracking floors worse than vanilla SGD in drift-dominated regimes.
STAMP adapter enables general time series foundation models to match specialized EEG foundation models on clinical classification tasks across 8 benchmarks while using few trainable parameters.
A FiLM-conditioned transformer masker on DAC codec latents performs text-guided sound separation with claimed efficiency, but the main comparison against AudioSep is confounded by asymmetric input processing.
Step-Audio 2 integrates a latent audio encoder, reasoning-centric reinforcement learning, and discrete audio token generation into language modeling to deliver state-of-the-art performance on audio understanding and conversational benchmarks.
A new open-source pipeline creates a 5,078-hour Russian speech dataset with prosody annotations, and VITS and SEMamba models trained on it outperform those trained on existing Russian corpora under equalized budgets.
Qwen-Audio trains a unified model on diverse audio and tasks with hierarchical tags to enable strong zero-shot performance on audio understanding benchmarks and multi-turn audio chat.
Introduces joint residual reweighting that disentangles speaker and joint residuals in CFG to improve speaker fidelity while preserving text accuracy in zero-shot TTS.
FusionSense uses server-side fusion learning, filter-out-safe labels, and edge compaction to enable runtime-adaptive multimodal sensing that cuts energy up to 33x while preserving task quality on RGB+Depth data.
WorldSpeech supplies 65k hours of multilingual aligned speech data across 76 languages and delivers 63.5% average relative WER reduction after fine-tuning ASR models on 11 typologically diverse languages.
citing papers explorer
-
ReasonAudio: A Benchmark for Evaluating Reasoning Beyond Matching in Text-Audio Retrieval
ReasonAudio benchmark reveals that state-of-the-art text-audio retrieval models struggle with reasoning tasks like negation and duration, and multimodal LLMs lose reasoning ability after contrastive fine-tuning.
-
Basic syntax from speech: Spontaneous concatenation in unsupervised deep neural networks
ciwGAN and fiwGAN models trained on isolated words spontaneously generate concatenated multi-word outputs and display early compositionality precursors.
-
Sarashina2.2-TTS: Tackling Kanji Polyphony in Japanese Speech Generation via Data Scaling and Targeted Data Synthesis
Sarashina2.2-TTS achieves SOTA kanji reading accuracy via data scaling and Joyo-kanji-targeted synthesis, introduces the Joyo Kanji Yomi Benchmark and Kana-CER metric, and shows stable cross-lingual performance.
-
Self-Stigma Is Not a Monolith, but Generic Empathy Is: Persona-Conditioned LLM Support for People Who Use Drugs
Four self-stigma personas identified via LPA on 1,174 Reddit users; persona-conditioned LLMs achieve targeted shifts but experts prefer generic empathy baselines.
-
PolySpeech-100: A Large-Scale Benchmark for Speech Understanding Across 100+ Languages and Dialects
PolySpeech-100 is a new benchmark for native-level speech comprehension across 110 linguistic variants that evaluates 22 models and reports E2E advantages on dialects, robustness gaps on low-resource languages, and degradation from Chain-of-Thought prompting.
-
Normative Networks for Source Separation via Local Plasticity and Dendritic Computation
Predictive Entropy Maximization performs competitive blind source separation using only local error-driven and Hebbian updates derived from a surrogate entropy objective with spectral error bounds.
-
Escaping the Linearity Trap: Manifold Detours for Black-Box Adversarial Attacks on Singing Audio Deepfake Detection
MARS is a transfer-based black-box attack that uses bi-level optimization on semantic and artifact anchors to escape the linearity trap and improve attack success rates on SSL-SVDD by up to 36%.
-
Transition-Matrix Regularization for Next Dialogue Act Prediction in Counselling Conversations
KL regularization aligning model predictions with empirical transition patterns improves macro-F1 by 9-42% in next dialogue act prediction on German counselling data and transfers to other datasets.
-
Listening Deepfake Detection: A New Perspective Beyond Speaking-Centric Forgery Analysis
Introduces the LDD task, ListenForge dataset built from five listening head generation methods, and MANet model that detects listening forgeries via motion inconsistencies guided by audio semantics.
-
DialToM: A Theory of Mind Benchmark for Forecasting State-Driven Dialogue Trajectories
LLMs excel at retrospective mental-state labeling on naturalistic dialogues but mostly fail a context-free prospective probe that maps isolated mental-state profiles to dialogue trajectories, despite expert 100% accuracy.
-
CleanCodec: Efficient and Robust Speech Tokenization via Perceptually Guided Encoding
CleanCodec reframes audio tokenization as a selective information bottleneck to encode only perceptually important features at 12.5 tokens per second, outperforming prior codecs in efficiency, speaker similarity, and intelligibility.
-
TB-AVA: Text as a Semantic Bridge for Audio-Visual Parameter Efficient Finetuning
TB-AVA uses text-mediated gated semantic modulation to enable efficient audio-visual alignment, achieving state-of-the-art results on AVE, AVS, and AVVP benchmarks.
-
DECKER: Domain-invariant Embedding for Cross-Keyboard Extraction and Recognition
DECKER is a domain-invariant four-stage framework (keyboard normalization, adversarial disentanglement, cross-keyboard contrastive alignment, acoustic style randomization) plus LLM post-processing that improves keystroke inference over baselines on the new HEAR dataset, especially in cross-keyboard
-
Emotion-Cause Pair Extraction in Conversations via Semantic Decoupling and Graph Alignment
SCALE disentangles emotion and cause representations in conversations and uses optimal transport for many-to-many global alignment, achieving SOTA on ECPEC benchmarks.
-
PRISM-CTG: A Foundation Model for Cardiotocography Analysis with Multi-View SSL
PRISM-CTG is the first large-scale foundation model for cardiotocography that uses multi-view self-supervised learning on unlabeled data to learn transferable representations, outperforming baselines on seven downstream tasks with external validation.
-
On the Provable Suboptimality of Momentum SGD in Nonstationary Stochastic Optimization
Momentum SGD pays a provable tracking penalty under distribution shift — lag growing as (1−β)⁻¹ and tracking floors worse than vanilla SGD in drift-dominated regimes.
-
STAMP: Spatial-Temporal Adapter with Multi-Head Pooling
STAMP adapter enables general time series foundation models to match specialized EEG foundation models on clinical classification tasks across 8 benchmarks while using few trainable parameters.
-
CodecSep: Prompt-Driven Universal Sound Separation on Neural Audio Codec Latents
A FiLM-conditioned transformer masker on DAC codec latents performs text-guided sound separation with claimed efficiency, but the main comparison against AudioSep is confounded by asymmetric input processing.
-
Step-Audio 2 Technical Report
Step-Audio 2 integrates a latent audio encoder, reasoning-centric reinforcement learning, and discrete audio token generation into language modeling to deliver state-of-the-art performance on audio understanding and conversational benchmarks.
-
Balalaika: Data-Centric, Prosody-Aware Annotation Pipeline for Russian Speech
A new open-source pipeline creates a 5,078-hour Russian speech dataset with prosody annotations, and VITS and SEMamba models trained on it outperform those trained on existing Russian corpora under equalized budgets.
-
Qwen-Audio: Advancing Universal Audio Understanding via Unified Large-Scale Audio-Language Models
Qwen-Audio trains a unified model on diverse audio and tasks with hierarchical tags to enable strong zero-shot performance on audio understanding benchmarks and multi-turn audio chat.
-
Joint Residual Reweighting for Classifier Free Guidance in Flow-Matching Zero-Shot TTS
Introduces joint residual reweighting that disentangles speaker and joint residuals in CFG to improve speaker fidelity while preserving text accuracy in zero-shot TTS.
-
FusionSense: Tri-Stage Near-Sensor Learning for Runtime-Adaptive Multimodal Edge Intelligence
FusionSense uses server-side fusion learning, filter-out-safe labels, and edge compaction to enable runtime-adaptive multimodal sensing that cuts energy up to 33x while preserving task quality on RGB+Depth data.
-
WorldSpeech: A Multilingual Speech Corpus from Around the World
WorldSpeech supplies 65k hours of multilingual aligned speech data across 76 languages and delivers 63.5% average relative WER reduction after fine-tuning ASR models on 11 typologically diverse languages.
-
Audio Spoof Detection with GaborNet
GaborNet replaces sinc functions with Gabor filters in raw-audio neural networks and is tested for audio spoof detection with augmentations in RawNet2 and RawGAT-ST.
-
R-FLoRA: Residual-Statistic-Gated Low-Rank Adaptation for Single-Image Face Morphing Attack Detection
R-FLoRA combines Laplacian residual statistics with a frozen vision transformer via gated low-rank adapters, residual fusion, and contrastive alignment to achieve better accuracy and generalization than prior single-image face morphing attack detectors.
-
Qwen3.5-Omni Technical Report
Qwen3.5-Omni scales an omnimodal model to hundreds of billions of parameters with 256k context, introduces ARIA for stable speech synthesis, and reports SOTA performance on 215 audio-visual benchmarks while adding multilingual and audio-visual coding capabilities.
-
Kimi-Audio Technical Report
Kimi-Audio is an open-source audio foundation model that achieves state-of-the-art results on speech recognition, audio understanding, question answering, and conversation after pre-training on more than 13 million hours of speech, sound, and music data.
-
Recursive QLSTM with Dynamic Variational Quantum Circuit Adaptation
The paper introduces Recursive QLSTM via metacore recursion, numerically tests variants on sequence lengths, and offers theoretical arguments for better temporal propagation.
-
RADAR Challenge 2026: Robust Audio Deepfake Recognition under Media Transformations
RADAR Challenge 2026 organizes a multilingual audio deepfake detection benchmark with media transformations, reporting participation from 33 development and 22 evaluation teams using EER metric.
-
Qwen2-Audio Technical Report
Qwen2-Audio is an open-source audio-language model that outperforms prior systems such as Gemini-1.5-pro on audio-centric instruction-following benchmarks after simplified prompt-based pre-training and expanded data.
-
Ensemble Deep Learning Approaches for AI-Altered Video Detection
Merging one audio and three video deepfake detectors with voting-based fusion gives ~70–73% accuracy on FakeAVCeleb, with the audio branch performing at chance in the wild.
-
Self-Modulating Quantum Fast-Weight Programmers for Efficient Adaptive Sequential Learning
Self-Modulating QFWP adds adaptive modulation to quantum fast-weight updates and memory to improve stability and performance on sequential learning tasks.
-
Secure Password Generator Based on Secure Pseudo-Random Number Generator
A MAC-based PRNG for passwords is implemented and shown to meet NIST SP 800-90B entropy and IID criteria.
-
Quantum Adversarial Machine Learning: From Classical Adaptations to Quantum-Native Methods
A survey of quantum adversarial machine learning covering attacks, countermeasures, theoretical underpinnings, trends, and challenges.