LLM-Evolved Core War Programs Beat 96% of Human Players

MIT PhD student Akarsh Kumar told the MLST podcast how he used LLMs to evolve competing Core War assembly programs that beat 96% of human players. The experiment showed that evolution combined with a verifier far outperforms zero-shot code generation.

2026-10-08 ~ 2026-10-10 · 2 related posts

• Evolving Assembly Bots With LLMs: Core War Beats Zero-Shot, Says Akarsh Kumar — Machine Learning Street Talk · 2026-10-08
• LLMs evolve Core War warriors that beat or tie 96% of 294 human-written ones, says MLST podcast — Machine Learning Street Talk · 2026-10-10