spreadsheetbench-verified@1.0

A benchmark evaluating AI agents on real-world spreadsheet manipulation tasks (400 tasks from verified_400). Tasks involve Excel file manipulation including formula writing, data transformation, formatting, and conditional logic.

← Back to Registry

Run this task

CLI:

inspect eval inspect_harbor/spreadsheetbench_verified_1_0 --model openai/gpt-5

Python:

from inspect_ai import eval
from inspect_harbor import spreadsheetbench_verified_1_0

eval(spreadsheetbench_verified_1_0(), model="openai/gpt-5")

Dataset information

Harbor registry	spreadsheetbench-verified@1.0
Inspect task	spreadsheetbench_verified_1_0
Version	1.0
Samples	400
Paper	arxiv

See Task Parameters for the parameter set shared across all Harbor tasks.