ResearchDatasetsOfficeQAOfficeQAFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsOffice Application Question AnsweringOfficeQA held-out (test)72.1Score (%)59Question AnsweringOfficeQA82.5Accuracy25Question AnsweringOfficeQA 246 questions80.1Accuracy15Long-context reasoningOfficeQA57.14Accuracy10Agentic TaskOfficeQA36Accuracy5Question AnsweringOfficeQA (test)55.32Accuracy4