Few-shot Language Understanding on ARC-C, MMLU, TruthfulQA, Winogrande, HellaSwag, PIQA, and MathQA
26.79ARC-C ScoreFP16
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FP16BIT=16, Backbone=Qwen3-8B2026.07 | 26.79 | 26.37 | 45.79 | 50.43 | 31.94 | 62.02 | 24.46 | 38.26 | |
| GSRQBIT=2, Backbone=Qwen3-8B2026.07 | 23.89 | 29.91 | 45.65 | 53.91 | 29.47 | 73.23 | 29.18 | 40.75 | |
| GSRQBIT=0.75, Backbone=Qwen3-8B2026.07 | 22.95 | 25.87 | 50.51 | 51.85 | 26.86 | 65.4 | 22.48 | 37.99 | |
| VQLLMBIT=1, Backbone=Qwen3-8B2026.07 | 22.7 | 24.11 | 51.09 | 50.04 | 25.72 | 53.92 | 20.3 | 35.41 | |
| VQLLMBIT=2, Backbone=Qwen3-8B2026.07 | 22.18 | 25.88 | 51.64 | 50.67 | 25.47 | 56.86 | 22.35 | 36.44 | |
| GSRQBIT=1, Backbone=Qwen3-8B2026.07 | 21.5 | 26.19 | 48.82 | 54.14 | 27.47 | 66.38 | 24.52 | 38.43 |