Dataset · harborframework
Large input files for Terminal-Bench-Science
harborframework/terminal-bench-science-lfs
Mirror of heavy files for scientific tasks of the Terminal-Bench-Science benchmark excluded from git.
What's inside
The repository stores large input files that are pulled during Docker container builds for Terminal-Bench-Science benchmark tasks. All data is sorted into specific task directories and verified via checksums. The author explicitly warns that these are benchmark materials and must be excluded from training sets.
What you can do with it
- Local deployment and running agents on Terminal-Bench-Science benchmark tasks
- Filtering training corpora for test data leakage (decontamination)
- Downloads this month
- 365.1K
- Likes
- 11
How to load it
from datasets import load_dataset
ds = load_dataset("harborframework/terminal-bench-science-lfs")