OpenPI2.0: An Improved Dataset for Entity Tracking in Texts
This work addresses the challenge of understanding dynamic text for AI models, but it is incremental as it builds on an existing dataset with refinements.
The authors tackled the problem of tracking entity state changes in text by improving the OpenPI dataset to OpenPI2.0 with canonicalized entities and salience annotations, finding that current language models perform poorly in this fairer evaluation and that using salient entity changes as prompts improves downstream tasks like question answering and planning.
Much text describes a changing world (e.g., procedures, stories, newswires), and understanding them requires tracking how entities change. An earlier dataset, OpenPI, provided crowdsourced annotations of entity state changes in text. However, a major limitation was that those annotations were free-form and did not identify salient changes, hampering model evaluation. To overcome these limitations, we present an improved dataset, OpenPI2.0, where entities and attributes are fully canonicalized and additional entity salience annotations are added. On our fairer evaluation setting, we find that current state-of-the-art language models are far from competent. We also show that using state changes of salient entities as a chain-of-thought prompt, downstream performance is improved on tasks such as question answering and classical planning, outperforming the setting involving all related entities indiscriminately. We offer OpenPI2.0 for the continued development of models that can understand the dynamics of entities in text.