When do neural networks learn world models?
Humans develop world models that capture the underlying generation process of data. Whether neural networks can learn similar world models remains an open problem. In this work, we provide the first theoretical results for this problem, showing that in a multi-task setting, models with a low-degree bias provably recover latent data-generating variables under mild assumptions -- even if proxy tasks involve complex, non-linear functions of the latents. However, such recovery is also sensitive to model architecture. Our analysis leverages Boolean models of task solutions via the Fourier-Walsh transform and introduces new techniques for analyzing invertible Boolean transforms, which may be of independent interest. We illustrate the algorithmic implications of our results and connect them to related research areas, including self-supervised learning, out-of-distribution generalization, and the linear representation hypothesis in large language models.
Code (0)
등록된 구현이 없습니다.
Tasks
Out-of-Distribution GeneralizationSelf-Supervised LearningSimilar Papers 제목 키워드 기반
Novelty Detection in Reinforcement Learning with World Models
Reinforcement learning (RL) using world models has found significant recent successes. However, when a sudden change to world mechanics or properties occurs then agent performance and reliability can dramatically decline…
Decision MakingNovelty Detectionreinforcement-learningReinforcement Learning+1Sample Efficient Robot Learning with Structured World Models
Reinforcement learning has been demonstrated as a flexible and effective approach for learning a range of continuous control tasks, such as those used by robots to manipulate objects in their environment. But in robotics…
continuous-controlContinuous ControlA Retrospective on the Robot Air Hockey Challenge: Benchmarking Robust, Reliable, and Safe Learning Techniques for Real-world Robotics
Machine learning methods have a groundbreaking impact in many application domains, but their application on real robotic platforms is still limited. Despite the many challenges associated with combining machine learning …
BenchmarkingWhen and How Much to Imagine: Adaptive Test-Time Scaling with World Models for Visual Spatial Reasoning
Despite rapid progress in MLLMs, visual spatial reasoning remains unreliable when correct answers depend on how a scene would appear under unseen or alternative viewpoints. Recent work addresses this by augmenting reason…
Spatial ReasoningChanging Simplistic Worldviews
We study a Bayesian persuasion model with two-dimensional states of the world, in which the sender (she) and receiver (he) have heterogeneous prior beliefs and care about different dimensions. The receiver is a naive age…