HuBERT

Meta AI

Structures found in this family (5)

Linear Direction, Linear Separability, Representational-similarity trajectory across depth, Dimensional collapse, Linear Subspace

By model (2)

HuBERT-base · 95M

Linear Direction, Linear Separability, Representational-similarity trajectory across depth, Dimensional collapse, Linear Subspace

HuBERT Large (LibriLight 60k)

Linear Direction, Linear Subspace

Observations (7)

TopK SAE features on real Whisper and HuBERT encoders are seed-stable and sparsely erasable, and steering them causally cuts Whisper false-speech detections by 70%

Phonological features are linear vector-offset directions in speech models

Phonological contrast directions in frozen HuBERT-base embeddings collapse in separability with dysarthria severity, training-free and across languages

Real speech-SSL models Wav2Vec2, WavLM, HuBERT, and Data2Vec show distinct layer-wise entropy and trajectory-curvature regimes, with Wav2Vec2 undergoing a late-layer entropy collapse

Audio-model SAEs raise DCI completeness while preserving informativeness

Speech models encode neighbor-phone info in orthogonal positional subspaces

Speech SSL features linearly recover universal articulatory kinematics

Papers

AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders (2026), [b]=[d]-[t]+[p]: Self-Supervised Speech Models Discover Phonological Vector Arithmetic (2026), Training-Free Cross-Lingual Dysarthria Severity Assessment via Phonological Subspace Analysis in Self-Supervised Speech Representations (2026), InsideSSL: Understanding Self-Supervised Speech Representations using a Model-Centric Perspective (2026), Sparse Autoencoders Make Audio Foundation Models More Explainable (2025), Self-Supervised Speech Models Encode Phonetic Context via Position-dependent Orthogonal Subspaces (2026), Self-Supervised Models of Speech Infer Universal Articulatory Kinematics (2023)