Skip to content

Dataset · harborframework

Terminal scientific computing benchmark

harborframework/terminal-bench-science

Continuously updated benchmark of real computational workflows in natural, exact, and engineering sciences.

What's inside

The dataset mirrors the Terminal-Bench-Science GitHub repository and contains fewer than 1,000 text examples. Tasks cover real computational scenarios from biology, physics, Earth sciences, mathematics, and engineering. Benchmark versions are published as tags within a single repository.

What you can do with it

  • Evaluating AI agents in scientific development and terminal tasks
  • Testing code models on complex interdisciplinary pipelines
Size
up to 1K rows
Downloads this month
110.6K
Likes
22

How to load it

from datasets import load_dataset

ds = load_dataset("harborframework/terminal-bench-science")