Syntax Evaluation on BLiMP
84.61AccuracyGTCA
Evaluation Results
| Method | Links | |
|---|---|---|
| GTCABackbone Model=Llama-3-8B, Evaluation Protocol=0-shot2026.01 | 84.61 | |
| GTCABackbone Model=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.01 | 83.12 | |
| LoRA-onlyBackbone Model=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.01 | 80.87 | |
| LoRA-onlyBackbone Model=Llama-3-8B, Evaluation Protocol=0-shot2026.01 | 80.7 | |
| Direct-JointBackbone Model=Llama-3-8B, Evaluation Protocol=0-shot2026.01 | 80.36 | |
| Direct-JointBackbone Model=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.01 | 80.27 | |
| BackboneBackbone Model=Llama-3-8B, Evaluation Protocol=0-shot2026.01 | 79.95 | |
| BackboneBackbone Model=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.01 | 78.58 |