Latent Compositional Representations Improve Systematic Generalization in Grounded Question Answering
This addresses the challenge of systematic generalization for AI systems in grounded question answering, offering a novel method to improve robustness beyond in-distribution data.
The paper tackled the problem of poor systematic generalization in grounded question answering by proposing a model that induces latent compositional representations, achieving 96.1% accuracy on the CLOSURE dataset, significantly outperforming prior models on out-of-distribution examples.
Answering questions that involve multi-step reasoning requires decomposing them and using the answers of intermediate steps to reach the final answer. However, state-of-the-art models in grounded question answering often do not explicitly perform decomposition, leading to difficulties in generalization to out-of-distribution examples. In this work, we propose a model that computes a representation and denotation for all question spans in a bottom-up, compositional manner using a CKY-style parser. Our model induces latent trees, driven by end-to-end (the answer) supervision only. We show that this inductive bias towards tree structures dramatically improves systematic generalization to out-of-distribution examples, compared to strong baselines on an arithmetic expressions benchmark as well as on CLOSURE, a dataset that focuses on systematic generalization for grounded question answering. On this challenging dataset, our model reaches an accuracy of 96.1%, significantly higher than prior models that almost perfectly solve the task on a random, in-distribution split.