SpurAudio benchmark shows state-of-the-art few-shot audio classifiers suffer large performance drops when background correlations are disrupted, even in large pretrained models.
hub
wav2vec: Unsupervised Pre-training for Speech Recognition
17 Pith papers cite this work. Polarity classification is still indexing.
17
Pith papers citing it
hub tools
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 17roles
background 1polarities
background 1representative citing papers
Unimodal model representations converge to a relational structure captured by the Indra representation via V-enriched Yoneda embedding, which is unique and structure-preserving and improves cross-model and cross-modal robustness when instantiated with angular distance.
Spoof-SUPERB benchmark shows large-scale discriminative SSL models such as XLS-R, UniSpeech-SAT, and WavLM Large outperform others in audio deepfake detection and maintain robustness under acoustic degradations.
Flexformer learns attention kernels by treating spectral frequencies as trainable parameters in random Fourier feature-based linear attention, with stationary and nonstationary variants that outperform fixed-kernel baselines.
DYSCO jointly recovers latent trajectories and governing equations from noisy observations via multi-view contrastive learning, with theoretical guarantees up to affine indeterminacy.
A gated fusion of XLSR-53 and CORES features with energy margin and diversity losses reaches 97.6% ID accuracy and reduces FPR95 by 83.5% relative to the Interspeech 2025 baseline on MLAAD.
The paper introduces phoneme recognition using articulatory features as a proxy metric for evaluating articulatory speech synthesis quality from phonetic sequences.
LetsTalk combines a multimodal diffusion transformer, noise-regularized memory bank, deep compression autoencoder, and symbiotic/direct fusion schemes to achieve state-of-the-art quality and efficiency in long-duration talking video generation.
Adding register tokens to Vision Transformers eliminates high-norm background artifacts and raises state-of-the-art performance on dense visual prediction tasks.
wav2VOT shows wav2vec2 can estimate voice onset time and related stop consonant features with accuracy comparable to existing tools on unseen data and higher accuracy after fine-tuning.
Presents an end-to-end differentiable neural model for phoneme forced alignment that claims to outperform prior methods on English benchmarks and generalize to unseen languages.
HighSync is a diffusion-based lip synchronization system that operates natively at 512x512 resolution by eliminating data leakage to enforce genuine audio dependence and reports state-of-the-art results on quality and sync metrics.
Clustering DINO representations via K-means and LDA filters poisoned speech samples, reducing attack success rate from 99.75% to 0.25% at 10% poisoning level.
An adaptive cross-modal gating network improves depression detection from speech by selectively weighting sparse relevant segments across acoustic and textual modalities.
A survey that organizes audio SSL into five objective paradigms, relates their demands to architectural biases, and interprets downstream applications as tests of generalization.
EGI integrates four existing AI components for real-time multimodal emotion monitoring and feedback in simulated agile meetings, reporting 10% WER and improved self-awareness for Scrum Masters.
A survey provides a task-based formalization of meta-learning and meta-RL while chronicling algorithms that lead to DeepMind's Adaptive Agent.
citing papers explorer
-
SpurAudio: A Benchmark for Studying Shortcut Learning in Few-Shot Audio Classification
SpurAudio benchmark shows state-of-the-art few-shot audio classifiers suffer large performance drops when background correlations are disrupted, even in large pretrained models.
-
The Indra Representation Hypothesis for Multimodal Alignment
Unimodal model representations converge to a relational structure captured by the Indra representation via V-enriched Yoneda embedding, which is unique and structure-preserving and improves cross-model and cross-modal robustness when instantiated with angular distance.
-
A SUPERB-Style Benchmark of Self-Supervised Speech Models for Audio Deepfake Detection
Spoof-SUPERB benchmark shows large-scale discriminative SSL models such as XLS-R, UniSpeech-SAT, and WavLM Large outperform others in audio deepfake detection and maintain robustness under acoustic degradations.
-
Flexformer: Flexible Linear Transformer with Learnable Attention Kernel
Flexformer learns attention kernels by treating spectral frequencies as trainable parameters in random Fourier feature-based linear attention, with stationary and nonstationary variants that outperform fixed-kernel baselines.
-
Extracting Governing Equations from Latent Dynamics via Multi-View Contrastive Learning
DYSCO jointly recovers latent trajectories and governing equations from noisy observations via multi-view contrastive learning, with theoretical guarantees up to affine indeterminacy.
-
Dual-Branch Gated Fusion for Open-Set Audio Deepfake Source Tracing
A gated fusion of XLSR-53 and CORES features with energy margin and diversity losses reaches 97.6% ID accuracy and reduces FPR95 by 83.5% relative to the Interspeech 2025 baseline on MLAAD.
-
Evaluating Speech Articulation Synthesis with Articulatory Phoneme Recognition
The paper introduces phoneme recognition using articulatory features as a proxy metric for evaluating articulatory speech synthesis quality from phonetic sequences.
-
Multimodal Diffusion Transformer with Memory Bank for Scalable Long-Duration Talking Video Generation
LetsTalk combines a multimodal diffusion transformer, noise-regularized memory bank, deep compression autoencoder, and symbiotic/direct fusion schemes to achieve state-of-the-art quality and efficiency in long-duration talking video generation.
-
Vision Transformers Need Registers
Adding register tokens to Vision Transformers eliminates high-norm background artifacts and raises state-of-the-art performance on dense visual prediction tasks.
-
wav2VOT: Automatic estimation of voice onset time, closure duration, and burst realisation with wav2vec2
wav2VOT shows wav2vec2 can estimate voice onset time and related stop consonant features with accuracy comparable to existing tools on unseen data and higher accuracy after fine-tuning.
-
Fully Differentiable Neural Forced Alignment via Soft Dynamic Programming
Presents an end-to-end differentiable neural model for phoneme forced alignment that claims to outperform prior methods on English benchmarks and generalize to unseen languages.
-
HighSync: High-Quality Lip Synchronization via Latent Diffusion Models
HighSync is a diffusion-based lip synchronization system that operates natively at 512x512 resolution by eliminating data leakage to enforce genuine audio dependence and reports state-of-the-art results on quality and sync metrics.
-
Clustering Unsupervised Representations as Defense against Poisoning Attacks on Speech Commands Classification System
Clustering DINO representations via K-means and LDA filters poisoned speech samples, reducing attack success rate from 99.75% to 0.25% at 10% poisoning level.
-
Learning to Attend to Depression-Related Patterns: An Adaptive Cross-Modal Gating Network for Depression Detection
An adaptive cross-modal gating network improves depression detection from speech by selectively weighting sparse relevant segments across acoustic and textual modalities.
-
From Objectives to Applications: Aligning Architectural Biases in Audio Self-Supervised Learning
A survey that organizes audio SSL into five objective paradigms, relates their demands to architectural biases, and interprets downstream applications as tests of generalization.
-
EGI: A Multimodal Emotional AI Framework for Enhancing Scrum Master Real-time Self-Awareness
EGI integrates four existing AI components for real-time multimodal emotion monitoring and feedback in simulated agile meetings, reporting 10% WER and improved self-awareness for Scrum Masters.
-
Meta-Learning and Meta-Reinforcement Learning -- Tracing the Path towards DeepMind's Adaptive Agent
A survey provides a task-based formalization of meta-learning and meta-RL while chronicling algorithms that lead to DeepMind's Adaptive Agent.