SMAC-Talk: StarCraft Benchmark Tests LLM Agents Against Deceptive Allies

Article automatically generated from technical news.

SMAC-Talk extends StarCraft Multi-Agent Challenge with natural language communication, testing LLM agents against deceptive allies. Qwen3.5 models benchmarked; no model exceeds 72% win rate. Researchers released SMAC-Talk on June 2, 2026, a StarCraft benchmark that forces LLM agents to cooperate through natural language. The environment includes a deceptive communicator that actively lies to allies, testing whether agents can detect and overcome manipulation