ORCHARD: A Benchmark For Measuring Systematic Generalization of Multi-Hierarchical Reasoning
This work addresses the need for better benchmarks to assess reasoning capabilities in AI models, particularly for natural language processing, but it is incremental as it builds on prior evaluation frameworks like ListOps.
The authors tackled the problem of evaluating systematic generalization in neural sequence models by proposing ORCHARD, a diagnostic dataset for multi-hierarchical reasoning, and found that Transformer and LSTM models fail in systematic generalization, with Transformer performing no better than random when references between hierarchies increase.
The ability to reason with multiple hierarchical structures is an attractive and desirable property of sequential inductive biases for natural language processing. Do the state-of-the-art Transformers and LSTM architectures implicitly encode for these biases? To answer this, we propose ORCHARD, a diagnostic dataset for systematically evaluating hierarchical reasoning in state-of-the-art neural sequence models. While there have been prior evaluation frameworks such as ListOps or Logical Inference, our work presents a novel and more natural setting where our models learn to reason with multiple explicit hierarchical structures instead of only one, i.e., requiring the ability to do both long-term sequence memorizing, relational reasoning while reasoning with hierarchical structure. Consequently, backed by a set of rigorous experiments, we show that (1) Transformer and LSTM models surprisingly fail in systematic generalization, and (2) with increased references between hierarchies, Transformer performs no better than random.