Science
AtmosSci-Bench
Recorded language-model attempts on multiple-choice and open-ended atmospheric-science problems.
2,310items
54subjects
Dataset: CC-BY-NC-4.0, as declared in the upstream Croissant metadata. Code: MIT. Model outputs remain subject to the providers' terms identified in doc/sources_license.md.license
sciencedomain
reasoningdomain
textmodality
item-level responses released
Saturation status: No
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Score in [0, 1]