How to Motivate Your Dragon: Teaching Goal-Driven Agents to Speak and Act in Fantasy Worlds
This work addresses the challenge of developing interactive AI agents for complex, multi-modal tasks in fantasy worlds, representing an incremental advancement in agent-based systems.
The paper tackles the problem of creating goal-driven agents that can both act and communicate in fantasy text-game environments, by extending the LIGHT dataset with quests and introducing a reinforcement learning system that achieves zero-shot performance with human expert demonstrations.
We seek to create agents that both act and communicate with other agents in pursuit of a goal. Towards this end, we extend LIGHT (Urbanek et al. 2019) -- a large-scale crowd-sourced fantasy text-game -- with a dataset of quests. These contain natural language motivations paired with in-game goals and human demonstrations; completing a quest might require dialogue or actions (or both). We introduce a reinforcement learning system that (1) incorporates large-scale language modeling-based and commonsense reasoning-based pre-training to imbue the agent with relevant priors; and (2) leverages a factorized action space of action commands and dialogue, balancing between the two. We conduct zero-shot evaluations using held-out human expert demonstrations, showing that our agents are able to act consistently and talk naturally with respect to their motivations.