Assistive agents should make humans' lives easier.

A reinforcement learning (RL) agent is tasked with two fundamental, interdependent problems: exploration(howtodiscoverusefuldata),andcreditassignment(howtoincorporateit). The simplest way of estimating the value function is by averaging returns (futurediscountedsumsofrewards)startingfromtaking ainx.

Humans and animals have a rich and flexible understanding of the physical world, which enables them to infer the underlying dynamical trajectories of objects and events, plausible future states, and use that to plan and anticipate the consequences of actions.However, the neural mechanisms underlying these computations are unclear.We combine a goal-driven modeling approach with dense neurophysiological data and high-throughput human behavioral readouts that contain thousands of comparisons to directly impinge on this question.Specifically, we construct and evaluate several classes of sensory-cognitive networks to predict the future state of rich, ethologically-relevant environments, ranging from self-supervised end-to-end models with pixel-wise or object-slot objectives, to models that future predict in the latent space of purely static image-pretrained or dynamic video-pretrained foundation models.We find that ``scale is \emph{not} all you need'', and that many state-of-the-art machine learning models fail to perform well on our neural and behavioral benchmarks for future prediction.In fact, only one class of models matches these data well overall.We find that neural responses are currently best predicted by models trained to predict the future state of their environment in the \emph{latent} space of pretrained foundation models optimized for \emph{dynamic} scenes in a self-supervised manner.These models also approach the neurons' ability to predict the environmental state variables that are visually hidden from view, despite not being explicitly trained to do so.Finally, we find that not all foundation model latents are equal.Notably, models that future predict in the latent space of video foundation models that are optimized to support a \emph{diverse} range of egocentric sensorimotor tasks, reasonably match \emph{both} human behavioral error patterns and neural dynamics across all environmental scenarios that we were able to test.Overall, these findings suggest that the neural mechanisms and behaviors of primate mental simulation have strong inductive biases associated with them, and are thus far most consistent with being optimized to future predict on \emph{reusable} visual representations that are useful for Embodied AI more generally.

While deep reinforcement learning (RL) has been demonstrated effective in solving complex control tasks, sample efficiency remains a key challenge due to the large amounts of data required for remarkable performance. Existing research explores the application of representation learning for data-efficient RL, e.g., learning predictive representations by predicting long-term future states. However, many existing methods do not fully exploit the structural information inherent in sequential state signals, which can potentially improve the quality of long-term decision-making but is difficult to discern in the time domain. To tackle this problem, we propose State Sequences Prediction via Fourier Transform (SPF), a novel method that exploits the frequency domain of state sequences to extract the underlying patterns in time series data for learning expressive representations efficiently. Specifically, we theoretically analyze the existence of structural information in state sequences, which is closely related to policy performance and signal regularity, and then propose to predict the Fourier transform of infinite-step future state sequences to extract such information. One of the appealing features of SPF is that it is simple to implement while not requiring storage of infinite-step future states as prediction targets. Experiments demonstrate that the proposed method outperforms several state-of-the-art algorithms in terms of both sample efficiency and performance.

We develop a data-driven approach for runtime safety monitoring in flight testing, where pilots perform maneuvers on aircraft with uncertain parameters. Because safety violations can arise unexpectedly as a result of these uncertainties, pilots need clear, preemptive criteria to abort the maneuver in advance of safety violation. To solve this problem, we use offline stochastic trajectory simulation to learn a calibrated statistical model of the short-term safety risk facing pilots. We use flight testing as a motivating example for data-driven learning/monitoring of safety due to its inherent safety risk, uncertainty, and human-interaction. However, our approach consists of three broadly-applicable components: a model to predict future state from recent observations, a nearest neighbor model to classify the safety of the predicted state, and classifier calibration via conformal prediction. We evaluate our method on a flight dynamics model with uncertain parameters, demonstrating its ability to reliably identify unsafe scenarios, match theoretical guarantees, and outperform baseline approaches in preemptive classification of risk.

Weather prediction is a quintessential problem involving the forecasting of a complex, nonlinear, and chaotic high-dimensional dynamical system. This work introduces an efficient reduced-order modeling (ROM) framework for short-range weather prediction and investigates fundamental questions in dimensionality reduction and reduced order modeling of such systems. Unlike recent AI-driven models, which require extensive computational resources, our framework prioritizes efficiency while achieving reasonable accuracy. Specifically, a ResNet-based convolutional autoencoder augmented by block attention modules is developed to reduce the dimensionality of high-dimensional weather data. Subsequently, a linear operator is learned in the time-delayed embedding of the latent space to efficiently capture the dynamics. Using the ERA5 reanalysis dataset, we demonstrate that this framework performs well in-distribution as evidenced by effectively predicting weather patterns within training data periods. We also identify important limitations in generalizing to future states, particularly in maintaining prediction accuracy beyond the training window. Our analysis reveals that weather systems exhibit strong temporal correlations that can be effectively captured through linear operations in an appropriately constructed embedding space, and that projection error rather than inference error is the main bottleneck. These findings shed light on some key challenges in reduced-order modeling of chaotic systems and point toward opportunities for hybrid approaches that combine efficient reduced-order models as baselines with more sophisticated AI architectures, particularly for applications in long-term climate modeling where computational efficiency is paramount.

To address the dual challenges of inherent stochasticity and non-differentiable metrics in physical spatiotemporal forecasting, we propose Spatiotemporal Forecasting as Planning (SFP), a new paradigm grounded in Model-Based Reinforcement Learning. SFP constructs a novel Generative World Model to simulate diverse, high-fidelity future states, enabling an "imagination-based" environmental simulation. Within this framework, a base forecasting model acts as an agent, guided by a beam search-based planning algorithm that leverages non-differentiable domain metrics as reward signals to explore high-return future sequences. These identified high-reward candidates then serve as pseudo-labels to continuously optimize the agent's policy through iterative self-training, significantly reducing prediction error and demonstrating exceptional performance on critical domain metrics like capturing extreme events.