Representation Learning for Grounded Spatial Reasoning
This work addresses the challenge of interpreting spatial references in AI agents, which is incremental but provides strong performance gains in a specific domain.
The paper tackles the problem of spatial reasoning in simulated environments by learning world representations guided by instruction text, achieving a 45% reduction in goal localization error compared to state-of-the-art methods.
The interpretation of spatial references is highly contextual, requiring joint inference over both language and the environment. We consider the task of spatial reasoning in a simulated environment, where an agent can act and receive rewards. The proposed model learns a representation of the world steered by instruction text. This design allows for precise alignment of local neighborhoods with corresponding verbalizations, while also handling global references in the instructions. We train our model with reinforcement learning using a variant of generalized value iteration. The model outperforms state-of-the-art approaches on several metrics, yielding a 45% reduction in goal localization error.