Paraphrase Identification on PAWS-X 1.0 (test)
55.3Accuracy (de)TokAlign++
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TokAlign++Backbone=Pythia 1B, Tuning Stage=Initialization without any tuning2026.05 | 55.3 | 50.4 | 54.8 | 55.3 | 55.4 | |
| ZeTTBackbone=Pythia 6.9B, Tuning Stage=Initialization without any tuning2026.05 | 55.3 | 43.7 | 47.7 | 54.9 | 51.3 | |
| TokAlign++Backbone=Pythia 6.9B, Tuning Stage=Initialization without any tuning2026.05 | 55.1 | 50.4 | 55.9 | 55.2 | 55.2 | |
| TokAlign++Backbone=Pythia 1B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 54.5 | 52.7 | 55.9 | 54.9 | 55.3 | |
| TokAlignBackbone=Pythia 6.9B, Tuning Stage=Initialization without any tuning2026.05 | 54.5 | 47.8 | 55.1 | 53.1 | 52.9 | |
| FocusBackbone=Pythia 6.9B, Tuning Stage=Initialization without any tuning2026.05 | 54.4 | 51.4 | 49.1 | 55.2 | 50.9 | |
| TokAlignBackbone=Pythia 1B, Tuning Stage=Initialization without any tuning2026.05 | 52.1 | 54.4 | 53.4 | 55.2 | 50.3 | |
| TokAlignBackbone=Pythia 1B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 51.5 | 52.1 | 55.7 | 53.4 | 55.3 | |
| ZeTTBackbone=Pythia 1B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 51 | 50.2 | 49.4 | 54.8 | 55.2 | |
| TokAlign++Backbone=Pythia 6.9B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 51 | 48.7 | 55.9 | 55 | 55.3 | |
| ZeTTBackbone=Pythia 1B, Tuning Stage=Initialization without any tuning2026.05 | 50.2 | 53.9 | 52.1 | 55.1 | 49.8 | |
| FocusBackbone=Pythia 1B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 50.2 | 50.3 | 50.8 | 50 | 54.4 | |
| PythiaBackbone=Pythia 1B, Tuning Stage=Vanilla (Zero-shot)2026.05 | 49.6 | 49.3 | 54.8 | 54.9 | 52.9 | |
| FocusBackbone=Pythia 6.9B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 48.8 | 46 | 51.4 | 52.2 | 54 | |
| FocusBackbone=Pythia 1B, Tuning Stage=Initialization without any tuning2026.05 | 48.6 | 50.4 | 45.7 | 44.8 | 47.3 | |
| ZeTTBackbone=Pythia 6.9B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 48.6 | 45.2 | 54.5 | 54.8 | 54.7 | |
| TokAlignBackbone=Pythia 6.9B, Tuning Stage=1k steps tuning on multilingual corpus2026.05 | 47.8 | 46.8 | 55.9 | 55.1 | 53.1 | |
| PythiaBackbone=Pythia 6.9B, Tuning Stage=Vanilla (Zero-shot)2026.05 | 43.8 | 40.2 | 50.2 | 54.2 | 50.2 |