MathArena has tested a few LLMs on the 2025 USAMO, achieving an average score of 2 points out of 42. Even I can score higher than that! Read the paper at https://arxiv.org/abs/2503.21934. Detailed info is available at https://matharena.ai/.
Maybe LLMs won't be taking over our jobs yet.
UPDATE: Gemini 2.5 pro apparently gets 10 points, which is better, but still not high enough to beat me.
Inspirational! I can't wait to see where your resilience and consistency takes you next!!!
The real question is can AI even qualify for USAMO (which is way harder than getting a 2 btw)
They score pretty high on computational because they only need to guess the answer instead of actually rigorously proving the answer. O3-mini seems to get an average of 12 and 14 on the AIME I and AIME II, respectively.
With some luck, maybe I can score higher than LLMs on the 2025 USAMO as well!
Just you wait, skin bags, just you wait
Auto comment: topic has been updated by bronze_coder (previous revision, new revision, compare).