Silver medal group's highest score, Google DeepMind breaks through the AI mathematical reasoning limit: solves 4 out of 6 international math Olympiad questions
The Google DeepMind team yesterday (July 25) released a blog post, stating that its developed AlphaProof and AlphaGeometry 2 models can solve advanced reasoning problems in mathematics, reaching the silver medal standard in answering difficult questions from the International Mathematical Olympiad.
AI translates the 6 IMO questions into mathematical language that the model can understand. AlphaProof solved two algebra problems and one number theory problem by determining the answers and proving their correctness, with one of the problems being the most difficult in the competition, which only 5 participants solved in this year's IMO. AlphaGeometry 2 proved geometry problems but did not solve 2 combinatorial problems.
According to IMO rules, each of the 6 questions can earn 7 points, with a maximum total score of 42 points. The DeepMind system ultimately scored 28 points, achieving full marks for each question, equivalent to the highest score in the silver medal group.