General
InfiBench
Per-attempt programming-question grades from the author-released InfiBench evaluation logs, including three human-answer comparators and the historical two-model development release.
369items
121subjects
Figshare evaluation release: CC BY 4.0. Both captured GitHub code repositories: Apache-2.0. Original question and answer sources retain their attribution requirements.license
codedomain
textmodality
item-level responses released
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Unbounded numeric score