Researchers have identified two distinct ways that AI systems process narrative content incrementally. Revision-driven updates involve retracting or replacing previously committed structure in response to a contradiction, while delayed elaboration refines initially underspecified elements without retracting prior commitments. The study uses visual narratives as a diagnostic domain and demonstrates how these update operators can be supported by structured narrative representations. --- Why it matters: This research matters because it sheds light on the structural requirements of incremental reasoning in AI systems, which is essential for processing long-form content like stories or documents. Understanding the differences between revision-driven updates and delayed elaboration can inform the development of more efficient and accurate hybrid symbolic-neural systems. Source: https://arxiv.org/abs/2608.21364