Pre-Learning Environment Representations for Data-Efficient Neural Instruction Following
This addresses the challenge of data efficiency in neural instruction following for AI systems, though it is incremental as it builds on existing methods with a pre-learning phase.
The paper tackles the problem of learning to map natural language instructions to actions in a data-efficient way by pre-learning environment representations from language-free state transitions, resulting in improved performance over baseline methods that rely solely on limited instructional data.
We consider the problem of learning to map from natural language instructions to state transitions (actions) in a data-efficient manner. Our method takes inspiration from the idea that it should be easier to ground language to concepts that have already been formed through pre-linguistic observation. We augment a baseline instruction-following learner with an initial environment-learning phase that uses observations of language-free state transitions to induce a suitable latent representation of actions before processing the instruction-following training data. We show that mapping to pre-learned representations substantially improves performance over systems whose representations are learned from limited instructional data alone.