Putnam-like problems in Lean

This repository contains Lean formalizations of Putnam-like problems. This repository is maintained by Eric Wieser and Henryk Michalewski.

Putnam-like is a benchmark crafted to probe the boundaries of mathematical reasoning in sophisticated AI systems. This collection offers university-level mathematical challenges, echoing the spirit of the Putnam Competition and similar student contests. These problems are intended to assess the depths of mathematical understanding, moving beyond superficial mathematical mimicry to demand a more profound and adaptable form of abstract thinking.

The repository at github.com/google-deepmind/eval_hub/tree/master/eval_hub/putnam_like contains natural language problem statements, grading rubrics, a more detailed README, and information about the math and Lean experts who contributed to this dataset.