Evolving Programmatic Skill Networks

Abstract

We study continual skill acquisition in open-ended embodied environments where an agent must construct, refine, and reuse an expanding library of executable skills. We introduce the Programmatic Skill Network (PSN), a framework in which skills are executable symbolic programs forming a compositional network that evolves through experience. PSN defines three core mechanisms instantiated via large language models: (1)REFLECT for structured fault localization over skill compositions, (2) progressive optimization with maturity-aware update gating that stabilizes reliable skills while maintaining plasticity for uncertain ones, and (3) canonical structural refactoring under rollback validation that maintains network compactness. We further show that PSN's learning dynamics exhibit structural parallels to neural network training. Experiments on MineDojo and Crafter demonstrate robust skill reuse, rapid adaptation, and strong generalization across open-ended task distributions.\footnote{We plan to open-source the code.

Repository Summary

PSN represents skills as executable symbolic programs in a compositional network that grows through reflection, maturity-aware optimization, and structural refactoring, and on Minecraft it unlocks diamond tools in 51 iterations on average vs. 102 for Voyager while also showing stronger Crafter learning curves.

Bibliographic Data

Title

Evolving Programmatic Skill Networks

Authors

Haochen Shi, Xingdi Yuan, Bang Liu

Publication date

2026/01/07

Identifier

arXiv:2601.03509

DOI

10.48550/arXiv.2601.03509

PDF size

3.1 MB