bespokelabs/autoresearch-exam

Coding

Science

AutoResearchExam measures an agent’s ability to improve and generalize on open ended machine learning research tasks. The dataset contains 29 tasks across machine learning research.

← Back to Registry

Run this task

CLI:

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

Python:

from inspect_ai import eval
from inspect_harbor import bespokelabs_autoresearch_exam

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

Dataset information

Harbor registry	bespokelabs/autoresearch-exam
Inspect task	`bespokelabs_autoresearch_exam`
Latest digest	sha256:3f2a6bbc4135e17b5f0df4416ef695fd5b999bd580cb5519f3000453460c31bd
Samples	29
Source	https://github.com/bespokelabsai/AutoResearchExam

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