Raycaster / evalsBack to SpreadsheetBench

SpreadsheetBench

41-47

1/1Pass

SpreadsheetBench task 41-47. Inspect the exact spreadsheet prompt, compare published model runs, open the agent response trace, and review grades for workbook editing.

Spreadsheet editingDual harness
ssb-41-47
SpreadsheetBench
2 models
Dual (parsed + original)
PromptWorkbook objective and answer cells
ResponseOpen each trace to inspect tool use and edits
GradeCompare score and pass/fail by model

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): 'CMSN!A1:D1529','CSTR!A1:D592,'EXPRO!A1:D8','OUT CAS!A1:O1529' Expected answer region(s): OUT CAS'!A2:C1529,'OUT CAS'!E2:G586,'OUT CAS'!I2:K13,'OUT CAS'!L2:O8 Expected answer sheet(s): OUT CAS Task: On OUT CAS tab, I want to aggregate and summarize by account types. See the headers in the OUT CAS tables for the account types. Go through the 3 other sheets and aggregate each item in the appropriate bucket on OUT CAS. Order them by item number (i.e., 000011 should be before 000012). PUR Purchase should be pretty much correct if you want to use that as the example. Additionally, I want to be able to insert new sheets before the OUT CAS sheet when needed, and also have the ability to clear the existing data in the OUT CAS sheet before performing the split.

Published trajectories

Agent runs on this task

Curated dual-harness runs. Best scored run per model.

ModelHarnessScoreResultLinks
GPT-5.4showcasedual1/1Pass
Gemini 3.1 Produal1/1Pass