mlsanityData to 14 Oct 2026

From Generalist to Specialist Representation

arXiv 2605.12733 ↗First posted 12 May 2026cs.LG cs.AI stat.MLcited by 0

Not on the map. Why?

Abstract (arXiv)

Given a generalist model, learning a task-relevant specialist representation is fundamental for downstream applications. Identifiability, the asymptotic guarantee of recovering the ground-truth representation, is critical because it sets the ultimate limit of any model, even with infinite data and computation. We study this problem in a completely nonparametric setting, without relying on interventions, parametric forms, or structural constraints. We first prove that the structure between time steps and tasks is identifiable in a fully unsupervised manner, even when sequences lack strict temporal dependence and may exhibit disconnections, and task assignments can follow arbitrarily complex and interleaving structures. We then prove that, within each time step, the task-relevant latent representation can be disentangled from the irrelevant part under a simple sparsity regularization, without any additional information or parametric constraints. Together, these results establish a hierarchical foundation: task structure is identifiable across time steps, and task-relevant latent representations are identifiable within each step. To our knowledge, each result provides a first general nonparametric identifiability guarantee, and together they mark a step toward provably moving from generalist to specialist models.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 36 of its 55 references that are in the map. First posted on arXiv 12 May 2026.

Builds on

36 of 55 references.

References · 36

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  3. 2018SlowFast Networks for Video Recognitioncited by 1,867
  4. 2018World Modelscited by 1,513
  5. 2022VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Trainingcited by 1,322
  6. 2019Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learningcited by 1,204
  7. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 941
  8. 2015Deep Learning and the Information Bottleneck Principlecited by 802
  9. 2018MINE: Mutual Information Neural Estimationcited by 683
  10. 2021Self-Supervised Learning with Data Augmentations Provably Isolates Content from Stylecited by 312
  11. 2018Nonlinear ICA Using Auxiliary Variables and Generalized Contrastive Learningcited by 282
  12. 2016Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICAcited by 273
24 more
  1. 2021Disentanglement via Mechanism Sparsity Regularization: A New Principle for Nonlinear ICAcited by 160
  2. 2022Weakly supervised causal representation learningcited by 140
  3. 2023Nonparametric Identifiability of Causal Representations from Unknown Interventionscited by 106
  4. 2021Learning Temporally Causal Latent Processes from General Temporal Datacited by 98
  5. 2021Independent mechanism analysis, a new concept?cited by 97
  6. 2022On the Identifiability of Nonlinear ICA: Sparsity and Beyondcited by 85
  7. 2022Identifiability of deep generative models without auxiliary informationcited by 70
  8. 2024Causal Representation Learning from Multiple Distributions: A General Settingcited by 68
  9. 2021Disentangling Identifiable Features from Noisy Data with Structured Nonlinear ICAcited by 64
  10. 2022Function Classes for Identifiable Nonlinear Independent Component Analysiscited by 53
  11. 2023Subspace Identification for Multi-Source Domain Adaptationcited by 37
  12. 2023Learning nonparametric latent causal graphs with unknown interventionscited by 34
  13. 2024Score-based Causal Representation Learning: Linear and General Transformationscited by 27
  14. 2021Decomposed Mutual Information Estimation for Contrastive Representation Learningcited by 25
  15. 2023Learning World Models with Identifiable Factorizationcited by 23
  16. 2023Learning Causal Representations from General Environments: Identifiability and Intrinsic Ambiguitycited by 15
  17. 2024SportsHHI: A Dataset for Human-Human Interaction Detection in Sports Videoscited by 15
  18. 2023Partial Identifiability for Domain Adaptationcited by 14
  19. 2025Modeling Open-World Cognition as On-Demand Synthesis of Probabilistic Modelscited by 14
  20. 2020C-MI-GAN : Estimation of Conditional Mutual Information using MinMax formulationcited by 10
  21. 2024Active Fine-Tuning of Multi-Task Policiescited by 9
  22. 2025When Do Neural Networks Learn World Models?cited by 8
  23. 2019Conditional Mutual Information Neural Estimatorcited by 7
  24. 2024Identifying Selections for Unsupervised Subtask Discoverycited by 4

Built on by

Not cited yet.