Commonsense Reasoning on XWinograd
80.7AccuracyBETR
Evaluation results
| Method | Links | |
|---|---|---|
| BETRTarget Setting=Single-Target2026.04 | 80.7 | |
| NAG_Llama-3.2-3BTarget Setting=Single-Target, Backbone=Llama-3.2-3B2026.04 | 80.6 | |
| NAG_SmolLM3-3BTarget Setting=Single-Target, Backbone=SmolLM3-3B2026.04 | 80.6 | |
| NAG_SmolLM3-3BTarget Setting=Multi-Target, Backbone=SmolLM3-3B2026.04 | 80.2 | |
| NAG_Qwen3-1.7BTarget Setting=Single-Target, Backbone=Qwen3-1.7B2026.04 | 80.1 | |
| NAG_Qwen3-1.7BTarget Setting=Multi-Target, Backbone=Qwen3-1.7B2026.04 | 79.9 | |
| NAG_Llama-3.2-3BTarget Setting=Multi-Target, Backbone=Llama-3.2-3B2026.04 | 79.9 | |
| Random2026.04 | 76.5 | |
| FineWeb-Edu2026.04 | 76.2 | |
| BETRTarget Setting=Multi-Target2026.04 | 76.1 | |
| 1B Model (HQ filtering)Filtering Strategy=Monolingual high-quality (HQ), Model Scale=1B2026.04 | 68.25 | |
| 1B Model (No filtering)Filtering Strategy=No filtering, Model Scale=1B2026.04 | 68.06 | |
| 1B Model (Multilingual classifier)Filtering Strategy=Multilingual (ML), Model Scale=1B2026.04 | 67.66 |