Mathematics
MathArena Platform
Published model attempts on ArXivMath, ArXivLean, BrokenArXiv and USAMO 2026, with original requests, full conversations and native grading scales.
1,414items
68subjects
Per-release CC-BY-SA-4.0 or CC-BY-NC-SA-4.0, as recorded in each captured Hugging Face dataset card; harness code is MIT.license
mathematicsdomain
reasoningdomain
textmodality
item-level responses released
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Item-specific grading scales