Hierarchical Decision Making by Generating and Following Natural Language Instructions
This work addresses the problem of coordinating large-scale, long-term actions in AI systems, offering a novel method for hierarchical planning with potential applications in robotics and game AI.
The paper tackles hierarchical decision making in complex real-time strategy games by using natural language instructions as latent plans, which are then executed by separate models. The approach significantly outperforms direct imitation of human actions, demonstrating the effectiveness of language's compositional structure for action representation.
We explore using latent natural language instructions as an expressive and compositional representation of complex actions for hierarchical decision making. Rather than directly selecting micro-actions, our agent first generates a latent plan in natural language, which is then executed by a separate model. We introduce a challenging real-time strategy game environment in which the actions of a large number of units must be coordinated across long time scales. We gather a dataset of 76 thousand pairs of instructions and executions from human play, and train instructor and executor models. Experiments show that models using natural language as a latent variable significantly outperform models that directly imitate human actions. The compositional structure of language proves crucial to its effectiveness for action representation. We also release our code, models and data.