Mathematical Reasoning on MATH500, AIME24, AIME25, AMC23
91.5MATH500 ScoreDS-Qwen-7B + LaneRoPE(NTK*)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DS-Qwen-7B + LaneRoPE(NTK*)Train.=KTO, N=22026.05 | 91.5 | 46.5 | 33.3 | 84.2 | 63.9 | |
| DS-Qwen-7B + LaneRoPE(NTK*)Train.=KTO, N=42026.05 | 90.7 | 46.3 | 33.6 | 85.9 | 64.1 | |
| DS-Qwen-7B + LaneRoPE(NTK*)Train.=KTO, N=12026.05 | 90.5 | 44.4 | 34.4 | 82.7 | 63 | |
| DS-Qwen-7B + BridgeTrain.=KTO, N=42026.05 | 89.4 | 40 | 33.2 | 85 | 61.9 | |
| DS-Qwen-7B + LaneRoPE(NTK*)Train.=SFT, N=22026.05 | 89 | 36 | 29.1 | 75.9 | 57.5 | |
| DS-Qwen-7B + BridgeTrain.=KTO, N=22026.05 | 88.8 | 40.5 | 29.2 | 81.4 | 60 | |
| DS-Qwen-7B + LaneRoPE(NTK*)Train.=SFT, N=42026.05 | 87.8 | 34.1 | 28.5 | 81.1 | 57.9 | |
| DS-Qwen-7B + LaneRoPE(NTK*)Train.=SFT, N=12026.05 | 87.3 | 32.6 | 28.5 | 76.7 | 56.3 | |
| DS-Qwen-1.5B + BridgeTrain.=KTO, N=42026.05 | 87 | 28.7 | 27.9 | 73 | 54.1 | |
| DS-Qwen-1.5B + BridgeTrain.=KTO, N=12026.05 | 86.9 | 32.1 | 26.6 | 74.2 | 54.9 | |
| DS-Qwen-1.5B + BridgeTrain.=KTO, N=22026.05 | 86.4 | 33.4 | 25.5 | 73.9 | 54.8 | |
| DS-Qwen-7BTrain.=-, N=12026.05 | 86.3 | 25.7 | 26.1 | 70.2 | 52.1 | |
| DS-Qwen-7B + BridgeTrain.=KTO, N=12026.05 | 85.5 | 39.4 | 29.4 | 82.8 | 59.3 | |
| DS-Qwen-7B + LaneRoPE(GT)Train.=-, N=22026.05 | 85 | 27.1 | 27.9 | 70.2 | 52.5 | |
| DS-Qwen-1.5B + LaneRoPE(NTK*)Train.=KTO, N=42026.05 | 84.6 | 26.3 | 25.6 | 68.7 | 51.3 | |
| DS-Qwen-1.5B + LaneRoPE(NTK*)Train.=KTO, N=12026.05 | 84.5 | 28.3 | 24.6 | 72.3 | 52.4 | |
| DS-Qwen-7B + LaneRoPE(NTK)Train.=-, N=22026.05 | 84.3 | 28.2 | 24.1 | 65.2 | 50.5 | |
| DS-Qwen-1.5B + LaneRoPE(NTK*)Train.=KTO, N=22026.05 | 83.9 | 26.3 | 25.2 | 71.3 | 51.7 | |
| DS-Qwen-1.5BTrain.=-, N=12026.05 | 79.2 | 19.4 | 19.1 | 56.5 | 43.5 | |
| DS-Qwen-7B + HogWildTrain.=-, N=22026.05 | 70.4 | 20.8 | 21.1 | 58 | 21.8 | |
| DS-Qwen-1.5B + HogWildTrain.=-, N=22026.05 | 58.6 | 7.5 | 6.6 | 43.3 | 27.2 | |
| DS-Qwen-1.5B + LaneRoPE(GT)Train.=-, N=22026.05 | 43.7 | 0 | 0 | 20.2 | 16 |