Datasetpublic

FineEnvs/SmolDataEnvs-harbor-test

📊 SmolDataEnvs: Harbor (test) 5.5K+ RL tasks for hill-climbing small models in code and data science. A 2B model on these tasks. Left: what it optimises. Right: 144 held-out tasks it never trains on. Two runs over the same 5,000 tasks: shuffled against a curriculum ordered easiest to hardest. The held-out benchmark: 250 tasks the model never trains on, weighted towards the hard end on purpose. This is the suite to quote a number from. Packaged in Harbor format… See the full description on the dataset page: https://huggingface.co/datasets/FineEnvs/SmolDataEnvs-harbor-test.

sourceHugging Facemitupdated 5h agoView on Hugging Face

0likes

21 commits on main

b1305955h ago

Card: drop em dashes

AdithyaSK

1b23b875h ago

Card: FineEnvs citation, install line for openenv[harbor]

AdithyaSK

937f5f75h ago

Card: full banner on top

AdithyaSK

81851de5h ago

Card GIF: hill climb only, no banner band

AdithyaSK

5a0b8a79h ago

README: the curves loop now carries the banner

AdithyaSK

4c2331213h ago

README: light-theme curves, split train | eval, shuffled vs curriculum

AdithyaSK

ef2a99013h ago

README: quote the tagline, add the training-curve loop

AdithyaSK

8e8ef6a13h ago

SmolDataEnvs: rename, new README, banner

AdithyaSK

a70c1883d ago

docs: link the Harbor Visualiser from the top of the card

AdithyaSK

e7a6ad33d ago

tags: declare as an RL environment dataset

AdithyaSK

291c8e511d ago

grader: stop emitting tool_efficiency, which was always null and destroyed correctness

AdithyaSK

89ae07f25d ago