Programmable Agents
This addresses the challenge of programmability and generalization in AI agents, offering a novel approach that is not explicitly incremental but introduces a new method for grounding language in RL.
The paper tackles the problem of enabling deep RL agents to execute declarative programs in a formal language, resulting in agents that generalize to new programs and objects not seen during training, with disentangled interpretable representations for zero-shot semantic tasks.
We build deep RL agents that execute declarative programs expressed in formal language. The agents learn to ground the terms in this language in their environment, and can generalize their behavior at test time to execute new programs that refer to objects that were not referenced during training. The agents develop disentangled interpretable representations that allow them to generalize to a wide variety of zero-shot semantic tasks.