Generative Question Answering on SQuAD Clean (test)
72.84Exact Match (EM)SDBN-p
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SDBN-pModel=Qwen-2.5-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 72.84 | — | |
| EDAModel=Qwen-2.5-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 59.48 | — | |
| NEFTuneModel=Qwen-2.5-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 58.12 | — | |
| VanillaModel=Qwen-2.5-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 58 | — | |
| SDBNBackbone=DeBERTa-v3, Training Samples=500, Average Runs=52026.06 | 57.8 | 72.2 | |
| VanillaBackbone=DeBERTa-v3, Training Samples=500, Average Runs=52026.06 | 55 | 69.8 | |
| NEFTuneBackbone=DeBERTa-v3, Training Samples=500, Average Runs=52026.06 | 55 | 69.6 | |
| FreeLBModel=Qwen-2.5-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 53.92 | — |