SpreadsheetBench
109-21
SpreadsheetBench task 109-21. Inspect the exact spreadsheet prompt, compare published model runs, open the agent response trace, and review grades for workbook editing.
Task prompt
What the agent was asked to do
You are solving a spreadsheet benchmark task in a real workbook. Objective: Produce the correct final workbook state for the expected answer region. What matters: - Only the values in the expected answer region will be graded. - The workbook is the answer. Instructions: 1. Read the workbook and inspect the relevant data region first. 2. Infer the required result for the provided workbook instance. 3. Write the final value(s) directly into the expected answer region. 4. Do not rely on prose, formulas in your chat response, pseudocode, or VBA as the answer unless the benchmark explicitly requires those to be written into cells. 5. If the natural-language task asks for a general method, formula, or macro, convert that into the concrete result needed for this workbook instance. 6. Keep your final text response short and only summarize the workbook cells you changed. Relevant data region(s): A1:X43 Expected answer region(s): 'Sheet1'!Z5:Z43 Expected answer sheet(s): Sheet1 Task: I need a macro to process a list, where groups are up to five people each. These people have point totals in column K. The task is to determine if an individual's point total is the highest in their group and indicate this with a 'Y' for yes or an 'N' for no in column Z. For individuals without a group (alone in their category), they should automatically receive a 'Y'. In case of a tie for the highest score within a group, all individuals with the tied highest score should get a 'Y'. The attached sample file represents a smaller portion of the actual data set which the macro will need to handle. The data is contained in columns A:X, and starts in row 5. People in the same group are in the same block (no empty rows between). No header for column Z.
Published trajectories
Agent runs on this task
Curated dual-harness runs. Best scored run per model.
| Model | Harness | Score | Result | Links |
|---|---|---|---|---|
| GPT-5.4showcase | dual | 1/1 | Pass | Share pagePublic trace |
| Gemini 3.1 Pro | dual | 1/1 | Pass | Share pagePublic trace |