DeepSeek-R1-Distill-Qwen-7B_eval_03-07-25_19-30_0981

# mlfoundations-dev/DeepSeek-R1-Distill-Qwen-7B_eval_03-07-25_19-30_0981 Precomputed model outputs for evaluation. ## Evaluation Results ### Summary | Metric | AIME24 | AIME25 | AMC23 | GPQADiamond | MATH500 | |--------|------|------|-----|-----------|-------| | Accuracy | 52.0 | 38.0 | 89.5 | 32.7 | 89.0 | ### AIME24 - **Average Accuracy**: 52.00% ± 4.77% - **Number of Runs**: 5 | Run | Accuracy | Questions Solved | Total Questions | |-----|----------|-----------------|----------------| | 1 | 50.00% | 15 | 30 | | 2 | 50.00% | 15 | 30 | | 3 | 53.33% | 16 | 30 | | 4 | 70.00% | 21 | 30 | | 5 | 36.67% | 11 | 30 | ### AIME25 - **Average Accuracy**: 38.00% ± 2.23% - **Number of Runs**: 5 | Run | Accuracy | Questions Solved | Total Questions | |-----|----------|-----------------|----------------| | 1 | 36.67% | 11 | 30 | | 2 | 43.33% | 13 | 30 | | 3 | 36.67% | 11 | 30 | | 4 | 30.00% | 9 | 30 | | 5 | 43.33% | 13 | 30 | ### AMC23 - **Average Accuracy**: 89.50% ± 0.84% - **Number of Runs**: 5 | Run | Accuracy | Questions Solved | Total Questions | |-----|----------|-----------------|----------------| | 1 | 87.50% | 35 | 40 | | 2 | 92.50% | 37 | 40 | | 3 | 90.00% | 36 | 40 | | 4 | 90.00% | 36 | 40 | | 5 | 87.50% | 35 | 40 | ### GPQADiamond - **Average Accuracy**: 32.66% ± 6.48% - **Number of Runs**: 3 | Run | Accuracy | Questions Solved | Total Questions | |-----|----------|-----------------|----------------| | 1 | 23.74% | 47 | 198 | | 2 | 25.76% | 51 | 198 | | 3 | 48.48% | 96 | 198 | ### MATH500 - **Accuracy**: 89.00% | Accuracy | Questions Solved | Total Questions | |----------|-----------------|----------------| | 89.00% | 445 | 500 |# mlfoundations-dev/DeepSeek-R1-Distill-Qwen-7B_eval_03-07-25_19-30_0981 用于评估的预计算模型输出。 ## 评估结果 ### 评估概览 | 指标 | AIME24 | AIME25 | AMC23 | GPQADiamond | MATH500 | |--------|------|------|-----|-----------|-------| | 准确率 | 52.0 | 38.0 | 89.5 | 32.7 | 89.0 | ### AIME24 - **平均准确率**：52.00% ± 4.77% - **测试轮次**：5 | 轮次 | 准确率 | 答对题目数 | 总题目数 | |-----|----------|-----------------|----------------| | 1 | 50.00% | 15 | 30 | | 2 | 50.00% | 15 | 30 | | 3 | 53.33% | 16 | 30 | | 4 | 70.00% | 21 | 30 | | 5 | 36.67% | 11 | 30 | ### AIME25 - **平均准确率**：38.00% ± 2.23% - **测试轮次**：5 | 轮次 | 准确率 | 答对题目数 | 总题目数 | |-----|----------|-----------------|----------------| | 1 | 36.67% | 11 | 30 | | 2 | 43.33% | 13 | 30 | | 3 | 36.67% | 11 | 30 | | 4 | 30.00% | 9 | 30 | | 5 | 43.33% | 13 | 30 | ### AMC23 - **平均准确率**：89.50% ± 0.84% - **测试轮次**：5 | 轮次 | 准确率 | 答对题目数 | 总题目数 | |-----|----------|-----------------|----------------| | 1 | 87.50% | 35 | 40 | | 2 | 92.50% | 37 | 40 | | 3 | 90.00% | 36 | 40 | | 4 | 90.00% | 36 | 40 | | 5 | 87.50% | 35 | 40 | ### GPQADiamond - **平均准确率**：32.66% ± 6.48% - **测试轮次**：3 | 轮次 | 准确率 | 答对题目数 | 总题目数 | |-----|----------|-----------------|----------------| | 1 | 23.74% | 47 | 198 | | 2 | 25.76% | 51 | 198 | | 3 | 48.48% | 96 | 198 | ### MATH500 - **准确率**：89.00% | 准确率 | 答对题目数 | 总题目数 | |----------|-----------------|----------------| | 89.00% | 445 | 500 |

搜集汇总

背景与挑战

背景概述

该数据集为DeepSeek-R1-Distill-Qwen-7B模型在多个评估基准上的预计算输出结果，包括AIME24、AIME25、AMC23、GPQADiamond和MATH500的准确率指标。数据集由mlfoundations-dev发布，采用Apache License 2.0许可。