README.md

mlfoundations-dev/fig1_all_s1_eval_636d

Precomputed model outputs for evaluation.

Evaluation Results

Summary

Metric	AIME24	AMC23	MATH500	MMLUPro	JEEBench	GPQADiamond	LiveCodeBench	CodeElo	CodeForces
Accuracy	20.7	60.2	77.0	28.8	41.9	41.4	23.2	3.9	5.3

AIME24

• Average Accuracy: 20.67% ± 2.04%
• Number of Runs: 10

Run	Accuracy	Questions Solved	Total Questions
1	30.00%	9	30
2	16.67%	5	30
3	30.00%	9	30
4	16.67%	5	30
5	26.67%	8	30
6	20.00%	6	30
7	13.33%	4	30
8	20.00%	6	30
9	10.00%	3	30
10	23.33%	7	30

AMC23

• Average Accuracy: 60.25% ± 1.30%
• Number of Runs: 10

Run	Accuracy	Questions Solved	Total Questions
1	65.00%	26	40
2	57.50%	23	40
3	57.50%	23	40
4	67.50%	27	40
5	62.50%	25	40
6	60.00%	24	40
7	57.50%	23	40
8	60.00%	24	40
9	52.50%	21	40
10	62.50%	25	40

MATH500

• Accuracy: 77.00%

  Accuracy	Questions Solved	Total Questions
  77.00%	385	500

MMLUPro

• Average Accuracy: 28.80% ± 0.00%
• Number of Runs: 1

Run	Accuracy	Questions Solved	Total Questions
1	28.80%	144	500

JEEBench

• Average Accuracy: 41.91% ± 0.54%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	40.58%	209.0	515
2	42.52%	219.0	515
3	42.62%	219.5	515

GPQADiamond

• Average Accuracy: 41.41% ± 1.09%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	40.91%	81	198
2	39.39%	78	198
3	43.94%	87	198

LiveCodeBench

• Average Accuracy: 23.22% ± 0.43%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	22.70%	116	511
2	22.90%	117	511
3	24.07%	123	511

CodeElo

• Average Accuracy: 3.92% ± 0.74%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	2.56%	10	391
2	4.09%	16	391
3	5.12%	20	391

CodeForces

• Average Accuracy: 5.30% ± 0.34%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	5.74%	26	453
2	4.64%	21	453
3	5.52%	25	453