Mathematical Reasoning on AQUA-RAT standard (test)
40.16ACC (%)Llama2-70B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Llama2-70BOutput Length=1002024.05 | 40.16 | — | — | 9,224.76 | — | |
| GKT (Llama2-70B -> Llama2-7B)Teacher Model=Llama2-70B, Student Model=Llama2-7B, Output Length=20->1002024.05 | 30.31 | 5.51 | 3.14 | 1,070.02 | 8.62 | |
| Llama2-7BOutput Length=1002024.05 | 27.17 | — | — | 707.02 | 13.05 |