Question Answering on TweetQA (eval)
80.81Clean F1SDBN-p
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SDBN-pBackbone=LLaMA-2-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 80.81 | 65.55 | 64.15 | |
| FreeLBBackbone=LLaMA-2-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 76.57 | 60.59 | 60.79 | |
| EDABackbone=LLaMA-2-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 70.02 | 53.14 | 55.04 | |
| NEFTuneBackbone=LLaMA-2-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 69.34 | 55.13 | 54.3 | |
| VanillaBackbone=LLaMA-2-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 68.09 | 51.56 | 56.06 | |
| SMARTBackbone=LLaMA-2-7B, PEFT Method=LoRA, Training Samples=2002026.06 | 66.71 | 49.54 | 51.13 |