Search papers, labs, and topics across Lattice.
This study investigates the discrepancies between the semantic content of Office Open XML (OOXML) documents as rendered in Microsoft Office applications and their representations when ingested by large language models (LLMs). The authors identify and characterize 21 distinct "evidence forks" that lead to divergent views of the same OOXML file, revealing that LLMs can extract task-relevant information that is not visible in the Office editing canvas. The findings highlight a significant gap in the assumptions of semantic integrity in LLM pipelines, with 48-76% of trials revealing hidden facts that could mislead users relying on these models for accurate information retrieval.
LLMs can extract critical information from OOXML files that is invisible in Microsoft Office, revealing a hidden layer of semantic divergence that could compromise decision-making in financial and compliance contexts.
LLM pipelines increasingly ingest Office Open XML (OOXML) documents (Word, Excel, and PowerPoint files) as first-class evidence in financial, compliance, and retrieval-augmented workflows, implicitly assuming semantic integrity: that the evidence consumed by the model matches the content shown in the Microsoft Office suite editing canvas. We show that this assumption can fail in OOXML-to-LLM pipelines. The same specification-valid OOXML file can yield one evidentiary view in Microsoft Office and another when extracted for an LLM. Each view is treated as authoritative by its consumer, a condition we call plural ground truth. The ingestion contract rarely states which view and semantic roles become model evidence or preserves how that evidence was derived. We call the specification-grounded OOXML constructions that induce such divergence evidence forks. We systematically traverse and mine the OOXML specification and confirm 21 evidence forks across Excel, Word, and PowerPoint, spanning six dimensions of view construction. All 13 tools in our extraction panel emit evidence from at least one fork. We test four native-ingestion LLM APIs and seven web chatbots. Each test document carries a trap: a task-relevant fact exposed by extraction but not shown in Office. Across this 21-mechanism evaluation, the four APIs return the trap in 48--76% of trials. For 20 of 21 mechanisms, at least one of the eleven interfaces returns the trap. Our measurements further show that exposure is shaped upstream of the model by the ingestion path and extractor configuration. A source-level survey of sixteen popular open-source LLM projects further shows that default OOXML ingestion paths concentrate on affected extractor families.
