Mathematical Reasoning on GSM8K (Eval Accuracy)
73.8AccuracyLYNX
Evaluation Results
| Method | Links | |
|---|---|---|
| LYNXTotal Token=25.143M, Throughput=8,5882026.05 | 73.8 | |
| MSIFR + LYNXTotal Token=23.731M, Throughput=16,6052026.05 | 73.5 | |
| MSFIRTotal Token=24.611M, Throughput=15,8232026.05 | 73.4 | |
| TraditionalTotal Token=109.091M, Throughput=3,2912026.05 | 72 | |
| DEERTotal Token=26.709M, Throughput=8,0862026.05 | 67.1 |