From Generalist to Specialist Representation

2026-05-12

AI Summary

This study addresses the challenge of disentangling task-specific representations from a general-purpose model under an unsupervised, nonparametric setting without structural priors, while ensuring their identifiability. By analyzing the relational structure of tasks across time steps, the work establishes, for the first time within a nonparametric framework, the identifiability of task structures and demonstrates that sparsity-inducing regularization within individual time steps plays a critical role in disentangling task-relevant representations. The paper develops a hierarchical identifiability theory that formally bridges generic representations to task-specialized ones, thereby providing theoretical performance guarantees for downstream tasks.

Abstract

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.

Problem

• identifiability
• specialist representation
• nonparametric
• unsupervised learning
• disentanglement

Innovation

• identifiability
• nonparametric
• task-relevant representation
• sparsity regularization
• unsupervised disentanglement