importantSYS.SOURCE: StarSkirmish• 2026-10-16T14:18:44Z
StarSkirmish Bench v0.2 Evaluates LLMs in StarCraft Bot Development
StarSkirmish Bench v0.2 evaluates large language models' ability to develop StarCraft: Brood War Protoss bots through 1-hour coding sessions, measuring win rates against human-written bots. GPT-6 Astra leads the benchmark, while Gemini 4 Argon demonstrates strong performance despite lower reasoning effort compared to other flagships.
*** END OF TRANSMISSION ***