mlsanityData to 7 Oct 2026

Self-Supervised Speech Models Encode Phonetic Context via Position-dependent Orthogonal Subspaces

arXiv 2603.12642 ↗First posted 13 Mar 2026eess.AS cs.CL cs.LGcited by 7

Not on the map. Why?

Abstract (arXiv)

Transformer-based self-supervised speech models (S3Ms) are often described as contextualized, yet what this entails remains unclear. Here, we focus on how a single frame-level S3M representation can encode phones and their surrounding context. Prior work has shown that S3Ms represent phones compositionally; for example, phonological vectors such as voicing, bilabiality, and nasality vectors are superposed in the S3M representation of [m]. We extend this view by proposing that phonological information from a sequence of neighboring phones is also compositionally encoded in a single frame, such that vectors corresponding to previous, current, and next phones are superposed within a single frame-level representation. We show that this structure has several properties, including orthogonality between relative positions, and emergence of implicit phonetic boundaries. Together, our findings advance our understanding of context-dependent S3M representations.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Builds on 41 of its 51 references that are in the map. First posted on arXiv 13 Mar 2026.

Builds on

41 of 51 references.

References · 41

  1. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,853
  2. 2021HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Unitscited by 2,301
  3. 2021WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processingcited by 1,609
  4. 2019How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddingscited by 770
  5. 2022AudioLM: a Language Modeling Approach to Audio Generationcited by 611
  6. 2024Moshi: a speech-text foundation model for real-time dialoguecited by 536
  7. 2021Layer-wise Analysis of a Self-supervised Speech Representation Modelcited by 293
  8. 2022Comparative layer-wise analysis of self-supervised speech modelscited by 155
  9. 2024Self-Supervised Speech Representations are More Phonetic than Semanticcited by 70
  10. 2023What Do Self-Supervised Speech Models Know About Words?cited by 62
  11. 2023Analysing Discrete Self Supervised Speech Representation for Spoken Language Modelingcited by 47
  12. 2023Exploring Speech Recognition, Translation, and Understanding with Discrete Speech Units: A Comparative Studycited by 42
29 more
  1. 2022Word Discovery in Visually Grounded, Self-Supervised Speech Modelscited by 37
  2. 2025Discrete Audio Tokens: More Than a Survey!cited by 37
  3. 2023Probing self-supervised speech models for phonetic and phonemic information: a case study in aspirationcited by 35
  4. 2022Evidence of Vocal Tract Articulation in Self-Supervised Learning of Speechcited by 30
  5. 2024SyllableLM: Learning Coarse Semantic Units for Speech Language Modelscited by 24
  6. 2023Self-supervised Predictive Coding Models Encode Speaker and Phonetic Information in Orthogonal Subspacescited by 19
  7. 2024Human-like Linguistic Biases in Neural Speech Models: Phonetic Categorization and Phonotactic Constraints in Wav2Vec2.0cited by 19
  8. 2024Sylber: Syllabic Embedding Representation of Speech from Raw Audiocited by 19
  9. 2022Opening the Black Box of wav2vec Feature Encodercited by 18
  10. 2023Wave to Syntax: Probing spoken language models for syntaxcited by 18
  11. 2024Coding Speech through Vocal Tract Kinematicscited by 18
  12. 2023Self-Supervised Models of Speech Infer Universal Articulatory Kinematicscited by 16
  13. 2023An Information-Theoretic Analysis of Self-supervised Discrete Representations of Speechcited by 15
  14. 2025Leveraging Allophony in Self-Supervised Speech Models for Atypical Pronunciation Assessmentcited by 15
  15. 2023Understanding Probe Behaviors through Variational Bounds of Mutual Informationcited by 11
  16. 2026[b] = [d] - [t] + [p]: Self-supervised Speech Models Discover Phonological Vector Arithmeticcited by 11
  17. 2024Phonetic Segmentation of the UCLA Phonetics Lab Archivecited by 9
  18. 2022Silence is Sweeter Than Speech: Self-Supervised Model Using Silence to Store Speaker Informationcited by 7
  19. 2023PWESuite: Phonetic Word Embeddings and Tasks They Facilitatecited by 6
  20. 2024Fine-Grained and Interpretable Neural Speech Editingcited by 6
  21. 2017Learning weakly supervised multimodal phoneme embeddingscited by 5
  22. 2024Interface Design for Self-Supervised Speech Modelscited by 5
  23. 2025Effective Context in Neural Speech Modelscited by 5
  24. 2026ZeroSyl: Simple Zero-Resource Syllable Tokenization for Spoken Language Modelingcited by 5
  25. 2025On-device Streaming Discrete Speech Unitscited by 4
  26. 2025Iterative refinement, not training objective, makes HuBERT behave differently from wav2vec 2.0cited by 4
  27. 2025LinearVC: Linear transformations of self-supervised features through the lens of voice conversioncited by 3
  28. 2026Sylber 2.0: A Universal Syllable Embeddingcited by 2
  29. 2026Kanade: A Simple Disentangled Tokenizer for Spoken Language Modelingcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 7.

7 papers.

  1. 2026An Empirical Recipe for Universal Phone Recognitioncited by 5
  2. 2026Adapting Self-Supervised Speech Representations for Cross-lingual Dysarthria Detection in Parkinson's Diseasecited by 4
  3. 2026Training-Free Cross-Lingual Dysarthria Severity Assessment via Phonological Subspace Analysis in Self-Supervised Speech Representationscited by 2
  4. 2026Language Orthogonalization for Zero-Shot Cross-Lingual Audio Deepfake Detectioncited by 1
  5. 2026Phonological Subspace Collapse Is Aetiology-Specific and Cross-Lingually Stable: Evidence from 3,374 Speakerscited by 0
  6. 2026Phone Segmentation and Recognition through Phonological Activation Mappingcited by 0
  7. 2026Mitigating Accent-Language Confusion in Self-Supervised Speech Representations for Language Identificationcited by 0