Understanding Multimodal Procedural Knowledge by Sequencing Multimodal Instructional Manuals

The ability to sequence unordered events is evidence of comprehension and reasoning about real world tasks/procedures. It is essential for applications such as task planning and multisource instruction summarization. It often requires thorough understanding of temporal common sense and multimodal information, since these procedures are often conveyed by a combination of texts and images. While humans are capable of reasoning about and sequencing unordered procedural instructions, the extent to which the current machine learning methods possess such capability is still an open question. In this work, we benchmark models' capability of reasoning over and sequencing unordered multimodal instructions by curating datasets from online instructional manuals and collecting comprehensive human annotations. We find current state-of-the-art models not only perform significantly worse than humans but also seem incapable of efficiently utilizing multimodal information. To improve machines' performance on multimodal event sequencing, we propose sequence-aware pretraining techniques exploiting sequential alignment properties of both texts and images, resulting in >5% improvements on perfect match ratio. 6 We design an algorithm to compute the inter-annotator agreements (IAAs), see Append. Sec. B.3 for details. The IAAs for (multimodal, text-only, image-only) versions in Wiki-How is: (0.84, 0.82, 0.69), and (0.92, 0.87, 0.81) in RecipeQA. 7 The alternative order annotation IAAs for (multimodal, text-only, image-only) versions in WikiHow is: (0.73, 0.71, 0.78), and (0.79, 0.76, 0.79) in RecipeQA.