Human-Timescale Adaptation in an Open-Ended Task Space

TL;DR

This paper shows that large-scale reinforcement learning with meta-learning and attention-based memory enables agents to adapt quickly to new, open-ended 3D environments, resembling human-like in-context learning capabilities.

Contribution

It introduces a novel RL agent trained with meta-learning, attention-based memory, and curriculum learning, achieving rapid adaptation in open-ended 3D tasks.

Findings

Scaling laws relate network size, memory, and task diversity to performance.

The adaptive agent demonstrates hypothesis-driven exploration and efficient knowledge use.

Prompting with demonstrations enhances adaptation success.

Abstract

Foundation models have shown impressive adaptation and scalability in supervised and self-supervised learning problems, but so far these successes have not fully translated to reinforcement learning (RL). In this work, we demonstrate that training an RL agent at scale leads to a general in-context learning algorithm that can adapt to open-ended novel embodied 3D problems as quickly as humans. In a vast space of held-out environment dynamics, our adaptive agent (AdA) displays on-the-fly hypothesis-driven exploration, efficient exploitation of acquired knowledge, and can successfully be prompted with first-person demonstrations. Adaptation emerges from three ingredients: (1) meta-reinforcement learning across a vast, smooth and diverse task distribution, (2) a policy parameterised as a large-scale attention-based memory architecture, and (3) an effective automated curriculum that prioritises tasks at the frontier of an agent's capabilities. We demonstrate characteristic scaling laws with respect to network size, memory length, and richness of the training task distribution. We believe our results lay the foundation for increasingly general and adaptive RL agents that perform well across ever-larger open-ended domains.