README.md

mlfoundations-dev/reasoning_25_chat_eval_636d

Precomputed model outputs for evaluation.

Evaluation Results

Summary

Metric	AIME24	AMC23	MATH500	MMLUPro	JEEBench	GPQADiamond	LiveCodeBench	CodeElo	CodeForces
Accuracy	28.7	70.5	82.8	21.6	51.4	33.0	34.5	10.6	15.3

AIME24

• Average Accuracy: 28.67% ± 1.35%
• Number of Runs: 10

Run	Accuracy	Questions Solved	Total Questions
1	33.33%	10	30
2	30.00%	9	30
3	23.33%	7	30
4	33.33%	10	30
5	33.33%	10	30
6	30.00%	9	30
7	26.67%	8	30
8	20.00%	6	30
9	30.00%	9	30
10	26.67%	8	30

AMC23

• Average Accuracy: 70.50% ± 1.16%
• Number of Runs: 10

Run	Accuracy	Questions Solved	Total Questions
1	67.50%	27	40
2	72.50%	29	40
3	72.50%	29	40
4	70.00%	28	40
5	70.00%	28	40
6	67.50%	27	40
7	67.50%	27	40
8	80.00%	32	40
9	67.50%	27	40
10	70.00%	28	40

MATH500

• Accuracy: 82.80%

  Accuracy	Questions Solved	Total Questions
  82.80%	414	500

MMLUPro

• Average Accuracy: 21.60% ± 0.00%
• Number of Runs: 1

Run	Accuracy	Questions Solved	Total Questions
1	21.60%	108	500

JEEBench

• Average Accuracy: 51.44% ± 0.59%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	52.04%	268.0	515
2	52.28%	269.25	515
3	50.00%	257.5	515

GPQADiamond

• Average Accuracy: 33.00% ± 2.62%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	29.29%	58	198
2	39.39%	78	198
3	30.30%	60	198

LiveCodeBench

• Average Accuracy: 34.51% ± 0.94%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	35.03%	179	511
2	35.81%	183	511
3	32.68%	167	511

CodeElo

• Average Accuracy: 10.57% ± 0.56%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	11.00%	43	391
2	11.25%	44	391
3	9.46%	37	391

CodeForces

• Average Accuracy: 15.31% ± 0.52%
• Number of Runs: 3

Run	Accuracy	Questions Solved	Total Questions
1	14.35%	65	453
2	15.45%	70	453
3	16.11%	73	453