Natural Language Understanding on GLUE (MNLI, RTE, SST-2 Subset)
56MNLI AccuracyB4G
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| B4GModel=QWEN2.5-7B, Task=style personalization2026.03 | 56 | 50 | 80 | |
| B4GModel=QWEN2.5-7B, Task=safety enhancement2026.03 | 54 | 50 | 86 | |
| B4GModel=GEMMA-2-9B, Task=style personalization2026.03 | 46 | 50 | 97 | |
| B4GModel=QWEN2.5-7B, Task=access control2026.03 | 46 | 50 | 72 | |
| B4GModel=GEMMA-2-9B, Task=safety enhancement2026.03 | 45 | 50 | 98 | |
| B4GModel=GEMMA-2-9B, Task=access control2026.03 | 45 | 50 | 99 | |
| B4GModel=GEMMA-2-9B, Task=model identity2026.03 | 44 | 50 | 96 | |
| B4GModel=QWEN2.5-7B, Task=model identity2026.03 | 44 | 50 | 76 | |
| B4GModel=LLAMA2-13B, Task=access control2026.03 | 42 | 50 | 88 | |
| B4GModel=LLAMA3.1-8B, Task=style personalization2026.03 | 40 | 50 | 96 | |
| B4GModel=LLAMA3.1-8B, Task=access control2026.03 | 40 | 50 | 97 | |
| B4GModel=LLAMA3.1-8B, Task=safety enhancement2026.03 | 38 | 50 | 98 | |
| B4GModel=LLAMA3.1-8B, Task=model identity2026.03 | 38 | 50 | 96 | |
| B4GModel=LLAMA2-13B, Task=style personalization2026.03 | 38 | 50 | 96 | |
| B4GModel=LLAMA2-13B, Task=safety enhancement2026.03 | 36 | 50 | 98 | |
| B4GModel=LLAMA2-13B, Task=model identity2026.03 | 36 | 50 | 92 |