Disentangling Representations through Multi-task Learning

Pantelis Vafidis, Aman Bhargava, Antonio Rangel

Keywords: zero-shot generalization, disentanglement, interpretability, world models, multi-task learning, computational neuroscience, neuroAI, evidence accumulation, cognitive maps, continuous attractors, RNNs, transformers

TL;DR: We theoretically prove multi-task learning is guaranteed to lead to disentangled, generalizable representations in autoregressive models, and validate our theory on RNNs and transformers performing cognitive neuroscience evidence accumulation tasks.

Abstract: Intelligent perception and interaction with the world hinges on internal representations that capture its underlying structure ("disentangled" or "abstract" representations). Disentangled representations serve as world models, isolating latent factors of variation in the world along approximately orthogonal directions, thus facilitating feature-based generalization. We provide experimental and theoretical results guaranteeing the emergence of disentangled representations in agents that optimally solve multi-task evidence accumulation classification tasks, canonical in the neuroscience literature. The key conceptual finding is that, by producing accurate multi-task classification estimates, a system implicitly represents a set of coordinates specifying a disentangled representation of the underlying latent state of the data it receives. The theory provides conditions for the emergence of these representations in terms of noise, number of tasks, and evidence accumulation time, when the classification boundaries are affine in the latent space. Surprisingly, the theory also produces closed-form expressions for extracting the disentangled representation from the model's latent state $\mathbf Z(t)$. We experimentally validate these predictions in RNNs trained on multi-task classification, which learn disentangled representations in the form of continuous attractors, leading to zero-shot out-of-distribution (OOD) generalization in predicting latent factors. We demonstrate the robustness of our framework across autoregressive architectures, decision boundary geometries and in tasks requiring classification confidence estimation. We find that transformers are particularly suited for disentangling representations, which might explain their unique world understanding abilities. Overall, our framework establishes a formal link between competence at multiple tasks and the formation of disentangled, interpretable world models in both biological and artificial systems, and helps explain why ANNs often arrive at human-interpretable concepts, and how they both may acquire exceptional zero-shot generalization capabilities.

Primary Area: interpretability and explainable AI

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 11804