Instruction Following on IFEval (latest test)
59.52IF-pSG-Zero
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| G-ZeroBackbone=Llama-3.1-8B-Instruct, Training Method=G-Zero, Training Round=22026.05 | 59.52 | 66.35 | 70.38 | 75.78 | |
| G-ZeroBackbone=Llama-3.1-8B-Instruct, Training Method=G-Zero, Training Round=12026.05 | 58.59 | 66.35 | 69.78 | 75.54 | |
| base modelBackbone=Llama-3.1-8B-Instruct, Training Method=Base2026.05 | 58.41 | 65.8 | 69.42 | 75.3 | |
| R-ZeroBackbone=Llama-3.1-8B-Instruct, Training Method=R-Zero2026.05 | 56.94 | 62.47 | 66.74 | 72.93 | |
| G-ZeroBackbone=Qwen3-8B-Base, Training Method=G-Zero, Training Round=22026.05 | 43.81 | 50.83 | 57.92 | 63.43 | |
| base modelBackbone=Qwen3-8B-Base, Training Method=Base2026.05 | 43.07 | 50.28 | 56 | 61.75 | |
| G-ZeroBackbone=Qwen3-8B-Base, Training Method=G-Zero, Training Round=12026.05 | 42.33 | 50.83 | 56.24 | 62.71 | |
| R-ZeroBackbone=Qwen3-8B-Base, Training Method=R-Zero2026.05 | 37.56 | 42.79 | 54.12 | 57.37 |