Reproduce “Digital Red Queen: Adversarial Program Evolution in Core War with LLMs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Akarsh Kumar, Ryan Bahlous-Boldi, Prafull Sharma, Phillip Isola, Sebastian Risi, Yujin Tang, David Ha
The Digital Red Queen (DRQ) algorithm leverages Large Language Models to evolve self-modifying assembly-like programs in the Core War environment. This framework facilitates the study of open-ended adversarial dynamics, demonstrating how continuous coevolution leads to increasingly generalist and robust program strategies, exhibiting emergent phenotypic convergence alongside genotypic diversity.
Key insights
- Continuous adversarial pressure, even from a simple self-play setup, drives the evolution of increasingly robust and generalist program strategies in complex, competitive environments.
- Artificial evolutionary systems can exhibit phenotypic convergence (similar functional behaviors) while maintaining genotypic non-convergence (diverse underlying code implementations), akin to natural convergent evolution.
- LLMs provide effective priors for navigating sparse program search spaces, significantly accelerating the discovery of functional and adaptive programs within an evolutionary framework.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.