Vocabulary Difficulty Prediction on BEA Shared Task de 2026 (dev)
1.0873RMSEmultilingual-E5-large
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| multilingual-E5-largeModeling Strategy=Neural Fusion2026.06 | 1.0873 | 0.2407 | 0.8234 | 0.8414 | 0.6446 | |
| Textualised Feature2026.06 | 1.0961 | 0.2319 | 0.8252 | 0.8417 | 0.6466 | |
| BGE-M3Modeling Strategy=Neural Fusion2026.06 | 1.1457 | 0.1823 | 0.8135 | 0.826 | 0.6284 | |
| LaBSEModeling Strategy=Neural Fusion2026.06 | 1.1568 | 0.1712 | 0.8015 | 0.8139 | 0.6208 | |
| XLM-R-baseModeling Strategy=Neural Fusion2026.06 | 1.2211 | 0.1069 | 0.7885 | 0.8058 | 0.6093 | |
| Late fusion2026.06 | 1.2627 | 0.0653 | 0.7219 | 0.7365 | 0.5389 | |
| mBERTModeling Strategy=Neural Fusion2026.06 | 1.2693 | 0.0587 | 0.7481 | 0.764 | 0.5665 | |
| Baseline (closed)2026.06 | 1.328 | — | 0.753 | — | — |