🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

MathArena results for HMMT Feb 2025 are out, showing that high school math competitions are still far from being solved by frontier LLMs, with only o3-mini crossing the 50% mark!

How good are LLMs at producing constructive proofs? In our latest paper we introduce MathConstruct, a benchmark consisting of challenging olympiad-level problems where solution requires proof by construction.

Can LLMs actually solve hard math problems? Given the strong performance at AIME, we now go to the next tier: our MathArena team has conducted a detailed evaluation using the recent 2025 USA Math Olympiad. The results are… bad: all models scored less than 5%!

Big update to our MathArena USAMO evaluation: Gemini 2.5 Pro, which was released *the same day* as our benchmark, is the first model to achieve non-trivial amount of points (24.4%). The speed of progress is really mind-blowing.

Check out this recent work from our lab showing that benign-looking LLM's can hide backdoors that activate upon finetuning!