KAKA22/SpreadsheetBench-v2
SpreadsheetBench 2 Home Page SpreadsheetBench 2 is a benchmark for evaluating agents on end-to-end business spreadsheet workflows. Unlike existing benchmarks that focus on isolated manipulations, SpreadsheetBench 2 requires agents to (1) complete workflow-level goals through multi-step coordinated operations, (2) perform cross-sheet reasoningwithin complex multi-sheet workbooks, (3) produce deliverable-level outcomes including structured models, repaired spreadsheets, and… See the full description on the dataset page: https://huggingface.co/datasets/KAKA22/SpreadsheetBench-v2.
SpreadsheetBench 2
SpreadsheetBench 2 is a benchmark for evaluating agents on end-to-end business spreadsheet workflows. Unlike existing benchmarks that focus on isolated manipulations, SpreadsheetBench 2 requires agents to (1) complete workflow-level goals through multi-step coordinated operations, (2) perform cross-sheet reasoningwithin complex multi-sheet workbooks, (3) produce deliverable-level outcomes including structured models, repaired spreadsheets, and accurate visualizations.
Key Features
- End-to-End Workflows: Tasks are designed as self-contained, multi-stage objectives requiring sequences of coordinated spreadsheet operations, rather than atomic formula generation or local edits.
- Professional Financial Domain: Covers real business scenarios including financial modeling with multi-statement integration, systematic error debugging across 10 error types, and data visualization.
- Complex Workbook Structures: Tasks involve workbooks with numerous sheets per question andrequire extensive cell modifications. Debugging tasks require hundreds of cell edits on average, and financialmodeling tasks require over a thousand.