dabstep@1.0

DABstep: Data Agent Benchmark for Multi-step Reasoning. 450 tasks where agents analyze payment transaction data with Python/pandas to answer business questions.

← Back to Registry

Run this task

CLI:

inspect eval inspect_harbor/dabstep_1_0 --model openai/gpt-5

Python:

from inspect_ai import eval
from inspect_harbor import dabstep_1_0

eval(dabstep_1_0(), model="openai/gpt-5")

Dataset information

Harbor registry	dabstep@1.0
Inspect task	dabstep_1_0
Version	1.0
Samples	450
Paper	arxiv

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