ExcelAudit AI: Diving deeper into Spreadsheet Dataset’s Quality & Performance

ACL ARR 2026 January Submission1637 Authors

Keywords: Excel, benchmarks, quality assessment, analysis

Abstract: We often encounter new benchmarks for which manual analysis of diversity and coverage is time-consuming and subjective. This problem also translates while analyzing different model performances over benchmarks post evaluation, making it hard to pinpoint categorical scenarios where the model fails more often or which areas have now improved with the newer version. That's where ExcelAudit AI comes into place. It is an easy-to-use tool that generates reports and insights for analyzing and comparing both benchmarks and evaluated results. We provide demo use cases over SheetCopilot and SpreadsheetBench datasets and agents, with validation against human insights.

Paper Type: Short

Research Area: Information Extraction and Retrieval

Research Area Keywords: Excel,benchmarks,quality assessment,analysis

Contribution Types: Data analysis

Languages Studied: English

Submission Number: 1637