TacticCraft: Natural Language-Driven Tactical Adaptation for StarCraft II

TL;DR

This paper introduces TacticCraft, a method that enables StarCraft II AI agents to adapt their tactics based on high-level strategic directives using lightweight adapters, enhancing flexibility without sacrificing performance.

Contribution

It proposes an adapter-based approach for tactical conditioning of pre-trained StarCraft II agents, allowing strategic customization with minimal computational overhead.

Findings

Successfully modulates agent behavior across tactical dimensions

Maintains competitive performance while adapting tactics

Enables flexible tactical control with minimal overhead

Abstract

We present an adapter-based approach for tactical conditioning of StarCraft II AI agents. Current agents, while powerful, lack the ability to adapt their strategies based on high-level tactical directives. Our method freezes a pre-trained policy network (DI-Star) and attaches lightweight adapter modules to each action head, conditioned on a tactical tensor that encodes strategic preferences. By training these adapters with KL divergence constraints, we ensure the policy maintains core competencies while exhibiting tactical variations. Experimental results show our approach successfully modulates agent behavior across tactical dimensions including aggression, expansion patterns, and technology preferences, while maintaining competitive performance. Our method enables flexible tactical control with minimal computational overhead, offering practical strategy customization for complex real-time strategy games.