userbench/UserBench

Assistants

Coding

UserBench eval: 620 next-message tasks across 62 developers. Composer 2.5 classifies gold and predicted messages into multi-label acts; reward is set Jaccard/IoU.

← Back to Registry

Run this task

CLI:

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

Python:

from inspect_ai import eval
from inspect_harbor import userbench

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

Dataset information

Harbor registry	userbench/UserBench
Inspect task	`userbench`
Latest digest	sha256:5ae6956f943da5d0781cf835cd8025a0411160321bb048f12c77bde7aac46bda
Samples	620

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