AIME
AIME (American Invitational Mathematics Examination) is an evaluation benchmark for assessing AI mathematical reasoning capabilities, derived from problems used in the American Invitational Mathematics Examination. The benchmark tests models on complex reasoning and problem-solving across areas such as algebra, number theory, geometry, and combinatorics through competition-level mathematical problems requiring multi-step logical deduction. It serves as an important reference standard for measuring advanced mathematical reasoning in large language models.
Timeline (last 90 days)
AIME contains 15 problems with integer answers, emphasizing the integrated use of algebra, geometry, number theory, and combinatorics.
Experiments were conducted on two mathematical reasoning benchmarks, AIME and HMMT, and ST-OPSD improves accuracy over the original OPSD and other intervention methods.
The model achieved a perfect score on the AIME and HMMT February competition mathematics sets.
AIME (American Invitational Mathematics Examination) is an elite mathematics competition for high school students, characterized by problems with many reasoning steps and unique answers.
Saluki scored 78.0 on MBPP+, 67.5 on multi-step causal reasoning, and 80.0 on AIME, retaining 82%–85% of the original model's mathematical reasoning capability.
Model scores on competition mathematics (such as AIME and IMO) and code generation (such as Codeforces competition problems) have shown a near-vertical growth curve over the past year.
Mathematical reasoning benchmarks include MATH and AIME, which test symbolic derivation and multi-step calculation.
Some large language models have already achieved over 80% accuracy on competition-level problem sets such as MATH and AIME.
AIME has only 30 samples, and tiny score differences derived from running a small number of seeds are not statistically significant.
AIME has only 30 samples; claiming superiority by bolding a score of 80 against a baseline of 79 after running 4 seeds is statistically untenable.
All Facts (11)
AIME 满分为 150 分（15 题×10 分），已成为衡量 AI 模型数学推理能力的黄金标准之一
90%UnverifiedAIME（美国数学邀请赛）中人类顶尖高中生的平均得分仅在3-5分左右（满分15分）
70%UnverifiedAIME contains 15 problems with integer answers, emphasizing the integrated use of algebra, geometry, number theory, and combinatorics.
50%UnverifiedExperiments were conducted on two mathematical reasoning benchmarks, AIME and HMMT, and ST-OPSD improves accuracy over the original OPSD and other intervention methods.
50%UnverifiedAIME (American Invitational Mathematics Examination) is an elite mathematics competition for high school students, characterized by problems with many reasoning steps and unique answers.
50%UnverifiedSaluki scored 78.0 on MBPP+, 67.5 on multi-step causal reasoning, and 80.0 on AIME, retaining 82%–85% of the original model's mathematical reasoning capability.
50%UnverifiedModel scores on competition mathematics (such as AIME and IMO) and code generation (such as Codeforces competition problems) have shown a near-vertical growth curve over the past year.
50%UnverifiedMathematical reasoning benchmarks include MATH and AIME, which test symbolic derivation and multi-step calculation.
50%UnverifiedSome large language models have already achieved over 80% accuracy on competition-level problem sets such as MATH and AIME.
50%UnverifiedAIME has only 30 samples, and tiny score differences derived from running a small number of seeds are not statistically significant.
50%UnverifiedAIME has only 30 samples; claiming superiority by bolding a score of 80 against a baseline of 79 after running 4 seeds is statistically untenable.
50%