Natural Language Understanding on GLUE (Performance Scores)
76.8CoLA ScoreIndividual
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| IndividualBackbone=GPT-2, Protocol=Single-task fine-tuned2026.03 | 76.8 | 82.1 | 80.4 | 88.3 | 89.6 | 65.3 | 91.2 | 82 | |
| ACE-MergingBackbone=GPT-2, Protocol=Model Merging2026.03 | 70.3 | 69.9 | 71.8 | 76.72 | 79 | 62.5 | 88.5 | 74.1 | |
| Task ArithmeticBackbone=GPT-2, Protocol=Model Merging2026.03 | 68.7 | 68.6 | 69.6 | 70.5 | 81.8 | 47.3 | 83.6 | 70 | |
| Ties-MergingBackbone=GPT-2, Protocol=Model Merging2026.03 | 68.4 | 71.4 | 68.4 | 69.6 | 82.4 | 47.7 | 81.8 | 70 | |
| TSV-MBackbone=GPT-2, Protocol=Model Merging2026.03 | 65.6 | 75.4 | 58.6 | 64.4 | 86.2 | 55.6 | 85.7 | 70.2 | |
| RegMeanBackbone=GPT-2, Protocol=Model Merging2026.03 | 61.7 | 70.4 | 65.4 | 69.7 | 78.8 | 56 | 79.7 | 68.8 | |
| Weight AveragingBackbone=GPT-2, Protocol=Model Merging2026.03 | 55 | 55.1 | 51 | 57.6 | 76.7 | 44.8 | 52.5 | 56.1 | |
| Fisher MergingBackbone=GPT-2, Protocol=Model Merging2026.03 | 54.8 | 58 | 39.5 | 63.3 | 81.5 | 49.1 | 64.7 | 58.7 |