CODE-SHARP: Continuous Open-ended Discovery and Evolution of Skills as Hierarchical Reward Programs

TL;DR

CODE-SHARP introduces a framework that autonomously discovers and evolves skills as hierarchical reward programs using foundation models, enabling agents to learn complex tasks from scratch without human engineering.

Contribution

It presents a novel method for open-ended skill discovery and evolution using hierarchical reward programs encoded as Python scripts, reducing reliance on human-designed rewards.

Findings

Agents outperform previous methods by 6x and 2.6x in median performance on Craftax-Classic and XLand.

Agents trained with CODE-SHARP can craft tools and mine diamonds, demonstrating advanced capabilities.

Zero-shot generalization to long-horizon tasks on Craftax-Extended, matching ground-truth reward trained agents.

Abstract

A core quality of general intelligence is the ability to open-endedly expand and evolve its set of mastered skills autonomously. While recent Foundation Model (FM) driven approaches have shown promising results towards this goal, they typically rely on significant human-in-the-loop engineering, limiting their transferability to novel environments. To address this, we introduce Continuous Open-ended Discovery and Evolution of Skills as Hierarchical Reward Programs (CODE-SHARP), a framework that leverages FMs to open-endedly grow and evolve an archive of Python programs encoding skills to train a generalist agent policy entirely from scratch via reinforcement learning, directly from source code. These programs, termed Skills as Hierarchical Reward Programs (SHARPs), each encode a local success condition and a set of prerequisites delegated to previously discovered SHARPs. At runtime, SHARPs dynamically route the agent through their prerequisite chain based on the current state, rewarding each completion along the way, requiring the agent to learn only the marginal behaviour each new SHARP introduces, enabling efficient learning of long-horizon skills without any pre-defined rewards. On Craftax-Classic and XLand, agents trained fully autonomously by CODE-SHARP outperform previous works by 6x and 2.6x in median performance and are the only agents capable of crafting iron tools and mining diamonds. Scaled to Craftax-Extended, CODE-SHARP trains a generalist agent on over 90 discovered SHARPs, enabling the agent to solve challenging long-horizon tasks zero-shot, matching agents trained on ground-truth rewards.