UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark

TL;DR

This paper introduces UNICORN, a universal commonsense reasoning model, along with a new multitask benchmark RAINBOW and a cost equivalent curve evaluation method, demonstrating significant improvements and insights across multiple datasets.

Contribution

The paper presents UNICORN, a new state-of-the-art commonsense model, and introduces RAINBOW and the cost equivalent curve for evaluating generalization and transfer learning in commonsense AI.

Findings

Transfer learning generally improves performance with a specific recipe.

QA-based datasets transfer well with each other, unlike knowledge graphs.

Larger models benefit more from transfer learning.

Abstract

Commonsense AI has long been seen as a near impossible goal -- until recently. Now, research interest has sharply increased with an influx of new benchmarks and models. We propose two new ways to evaluate commonsense models, emphasizing their generality on new tasks and building on diverse, recently introduced benchmarks. First, we propose a new multitask benchmark, RAINBOW, to promote research on commonsense models that generalize well over multiple tasks and datasets. Second, we propose a novel evaluation, the cost equivalent curve, that sheds new insight on how the choice of source datasets, pretrained language models, and transfer learning methods impacts performance and data efficiency. We perform extensive experiments -- over 200 experiments encompassing 4800 models -- and report multiple valuable and sometimes surprising findings, e.g., that transfer almost always leads to better or equivalent performance if following a particular recipe, that QA-based commonsense datasets transfer well with each other, while commonsense knowledge graphs do not, and that perhaps counter-intuitively, larger models benefit more from transfer than smaller ones. Last but not least, we introduce a new universal commonsense reasoning model, UNICORN, that establishes new state-of-the-art performance across 8 popular commonsense benchmarks, aNLI (87.3%), CosmosQA (91.8%), HellaSWAG (93.9%), PIQA (90.1%), SocialIQa (83.2%), WinoGrande (86.6%), CycIC (94.0%) and CommonsenseQA (79.3%).