Natural Language Inference on MNLI (Accuracy, Normalized Accuracy)
96.2Normalized AccuracyEMR-Merging
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EMR-MergingBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 96.2 | — | |
| LinearBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 96 | — | |
| FR-MergingBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 95.5 | — | |
| SVDBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 95.4 | — | |
| NSCBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 94.9 | — | |
| TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 94.3 | — | |
| DARE-TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 94.3 | — | |
| RobustMergeBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 94.3 | — | |
| AdaMergingBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 94.3 | — | |
| TABackbone=LLAMA-3 8B, LoRA rank=162026.03 | 92.8 | — | |
| KnOTS-TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 92 | — | |
| KnOTS-DARE-TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 91.8 | — | |
| LoRA-LEGOBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 87.7 | — | |
| FinetunedBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | 90.8 |