Natural Language Understanding on GLUE (NLI, MRPC, QNLI, QQP Subset)
50.21NLI AccuracyPre-edited
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Pre-editedBackbone=LLaMA2, Number of Edits=8000, Edit Dataset=zsRE2026.01 | 50.21 | 81.52 | 49.92 | 53.35 | |
| HORSEBackbone=LLaMA2, Number of Edits=8000, Edit Dataset=zsRE2026.01 | 50.18 | 80.31 | 53.25 | 54.06 | |
| MALMENBackbone=LLaMA2, Number of Edits=8000, Edit Dataset=zsRE2026.01 | 47.48 | 34.66 | 51.03 | 27.38 | |
| PMETBackbone=LLaMA2, Number of Edits=8000, Edit Dataset=zsRE2026.01 | 46.13 | 42.52 | 49.59 | 53.47 | |
| MEMITBackbone=LLaMA2, Number of Edits=8000, Edit Dataset=zsRE2026.01 | 43.12 | 81.23 | 49.46 | 53.83 | |
| MENDBackbone=LLaMA2, Number of Edits=8000, Edit Dataset=zsRE2026.01 | 33.54 | 0 | 50.54 | 0 |