[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"gQS6mhBgsY":3},"# MiniF2F\n\n[![.github/workflows/ci.yml](https://github.com/google-deepmind/miniF2F/actions/workflows/ci.yml/badge.svg)](https://github.com/google-deepmind/miniF2F/actions/workflows/ci.yml)\n[![Gitpod Ready-to-Code](https://img.shields.io/badge/Gitpod-ready--to--code-blue?logo=gitpod)](https://gitpod.io/#https://github.com/google-deepmind/miniF2F)\n\nThis repository is a fork of\n[openai/miniF2F](https://github.com/openai/miniF2F), which is described in\n[MiniF2F: a cross-system benchmark for formal Olympiad-level mathematics](https://arxiv.org/abs/2109.00110).\n\nIt contains Lean 4 translations of the Lean 3 problems in the original,\ntranslated using [mathport](https://github.com/leanprover-community/mathport).\n\nCompared to the original, this:\n\n*   contains natural language docstrings taken from (best estimates of) the\n    source of the original problems, to make identification of misformalizations\n    easier.\n\n    These descriptions originate from some combination of:\n\n    *   The contest collection question archive on the\n        [AoPS forums](https://artofproblemsolving.com/community/c13_contest_collections)\n    *   The [MATH dataset](https://github.com/hendrycks/math)\n\n*   has many fewer misformalizations, with all known false statements removed,\n    and many statements strengthened to match the strength of the english\n    statement.\n\n*   simplifies the `Minif2fImport` strategy, instead importing all of mathlib.\n\nThis is the version of the benchmark on which AlphaProof is evaluated.\n\nThis is not an official Google product.\n",1777138841084]