Natural Language Understanding on GLUE (8-Task Subset: CoLA, SST-2, MRPC, QQP, STS-B, MNLI, QNLI, RTE)
69.8CoLA AccuracyPEML
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| PEMLBackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 69.8 | 96.7 | 84.8 | 90.5 | 89.2 | 96.4 | 95.2 | 89.9 | 89.1 | |
| HyperLoRABackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 68.8 | 96.4 | 92.6 | 87.9 | 92.8 | 89.5 | 94.2 | 89.1 | 88.9 | |
| LoRAHUBBackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 65.3 | 92.6 | 84.1 | 88.9 | 90.3 | 91.7 | 94.6 | 87.8 | 86.9 | |
| LoRAMoEBackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 63.7 | 96.6 | 85.5 | 90.6 | 92.2 | 91.2 | 95.7 | 92.1 | 88.4 | |
| UniPELTBackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 63.1 | 92.1 | 87.9 | 85.9 | 88.7 | 83.9 | 91.3 | 72.9 | 83.2 | |
| HyperFormerBackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 57.9 | 94.8 | 89.9 | 87.9 | 91.5 | 89.8 | 94.5 | 87.8 | 86.8 | |
| HyperPromptBackbone=LLaMA2-7B, Evaluation Protocol=standard multi-task settings2026.05 | 56.3 | 93.7 | 91.7 | 87.6 | 90.3 | 91.1 | 94.9 | 87.6 | 86.7 |