SpreadSheetBench
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Spreadsheet Manipulation and Reasoning | SpreadsheetBench held-out (test) | Score85 | 59 | |
| Spreadsheet Agent Skills | SpreadsheetBench (in-distribution) | Vrf Score48.33 | 30 | |
| Spreadsheet Reasoning | SpreadsheetBench (test) | Accuracy58.5 | 24 | |
| Spreadsheet Task Execution | SpreadsheetBench Full | Soft Score47.42 | 24 | |
| Spreadsheet Task Execution | SpreadsheetBench Verified (Held-out) | Verification Score70.5 | 24 | |
| Spreadsheet Understanding and Manipulation | SpreadsheetBench Soft Restriction | Cell-Level Accuracy75.56 | 24 | |
| Spreadsheet Understanding and Manipulation | SpreadsheetBench Hard Restriction | Cell-Level Accuracy66.67 | 21 | |
| Spreadsheet Reasoning | SpreadsheetBench Verified | Accuracy71.11 | 18 | |
| Program Repair | SpreadsheetBench (test) | Pass Rate84 | 18 | |
| Spreadsheet Manipulation | SpreadsheetBench | Pass@157.7 | 16 | |
| Coding Agent Capacity | SpreadsheetBench Verified | Score82.3 | 8 | |
| Spreadsheet manipulation and reasoning | SpreadsheetBench | Soft Accuracy28.8 | 8 | |
| Trace Analysis | SpreadsheetBench (SSB) | Coverage96 | 7 | |
| Spreadsheet Reasoning | SpreadsheetBench | Pass@150.5 | 7 | |
| Agentic Tool Use | SpreadSheetBench | Success Rate55.36 | 5 | |
| Agent Skill Performance | SpreadsheetBench | Vrf Score69.83 | 4 | |
| Spreadsheet Task Execution | SpreadsheetBench common tasks (test) | Pass Rate52 | 3 | |
| Scaffold modification performance | SpreadsheetBench (250 held-out tasks) | Mean Score57.4 | 3 | |
| Report Quality Evaluation | SpreadsheetBench (test) | Correctness4.278 | 2 |
Showing 19 of 19 rows