Natural Language Inference on MNLI GLUE (test)
86.56AccuracySynLearner
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SynLearnerBackbone Model=Qwen2.5-7B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 86.56 | 86.47 | |
| LLaMA3.1-70BBackbone Model=Qwen2.5-7B, Training Protocol=No Training (Direct Synthesis)2026.05 | 86.01 | 85.96 | |
| Qwen2.5-7BBackbone Model=Qwen2.5-7B, Training Protocol=No Training (Direct Synthesis)2026.05 | 86 | 85.94 | |
| LLaMA3.1-8BBackbone Model=Qwen2.5-7B, Training Protocol=No Training (Direct Synthesis)2026.05 | 85.98 | 85.96 | |
| DeepseekBackbone Model=Qwen2.5-7B, Training Protocol=No Training (Direct Synthesis)2026.05 | 85.81 | 85.67 | |
| InsCLBackbone Model=Qwen2.5-7B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 85.79 | 85.79 | |
| SEEKRBackbone Model=Qwen2.5-7B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 85.67 | 85.44 | |
| GORPBackbone Model=Qwen2.5-7B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 85.53 | 85.28 | |
| OriBackbone Model=Qwen2.5-7B, Training Protocol=No Training (Direct Synthesis)2026.05 | 85.5 | 85.37 | |
| FAPMBackbone Model=Qwen2.5-7B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 84.88 | 85.02 | |
| SynLearnerBackbone Model=LLaMA3.1-8B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 82.5 | 82.28 | |
| GORPBackbone Model=LLaMA3.1-8B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 80.91 | 80.49 | |
| DeepseekBackbone Model=LLaMA3.1-8B, Training Protocol=No Training (Direct Synthesis)2026.05 | 80.76 | 80.66 | |
| SEEKRBackbone Model=LLaMA3.1-8B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 80.06 | 79.99 | |
| InsCLBackbone Model=LLaMA3.1-8B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 79.53 | 79.1 | |
| LLaMA3.1-70BBackbone Model=LLaMA3.1-8B, Training Protocol=No Training (Direct Synthesis)2026.05 | 78.67 | 78.8 | |
| Qwen2.5-7BBackbone Model=LLaMA3.1-8B, Training Protocol=No Training (Direct Synthesis)2026.05 | 78.52 | 77.66 | |
| FAPMBackbone Model=LLaMA3.1-8B, Training Protocol=With Training (Continual Learning-Based Synthesis)2026.05 | 78.48 | 78.8 | |
| LLaMA3.1-8BBackbone Model=LLaMA3.1-8B, Training Protocol=No Training (Direct Synthesis)2026.05 | 76.92 | 78.62 | |
| OriBackbone Model=LLaMA3.1-8B, Training Protocol=No Training (Direct Synthesis)2026.05 | 73.18 | 73.87 |