Tool Use on T-Eval
80.7Overall ScorePROVE
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PROVEModel=Qwen3-8B2026.06 | 80.7 | 94.8 | 65.2 | 63 | 92.7 | 81.8 | 86.5 | — | |
| Qwen3-8B + PROVETraining Steps=350 GRPO steps2026.06 | 80.7 | — | — | — | — | — | — | 1.9 | |
| BaseModel=Qwen3-8B2026.06 | 78.8 | 98.8 | 64.7 | 64.2 | 88 | 77.8 | 79.2 | — | |
| PROVEModel=Qwen3-4B2026.06 | 78.2 | 98.5 | 68.7 | 63.7 | 90.6 | 78.7 | 68.8 | — | |
| Qwen3-4B + PROVETraining Steps=350 GRPO steps2026.06 | 78.2 | — | — | — | — | — | — | 1.5 | |
| PROVEModel=Granite-4.1-8B2026.06 | 77.9 | 94.7 | 69.7 | 63.7 | 91.7 | 76.9 | 70.8 | — | |
| Granite-4.1-8B + PROVETraining Steps=350 GRPO steps2026.06 | 77.9 | — | — | — | — | — | — | 4.4 | |
| BaseModel=Qwen3-4B2026.06 | 76.7 | 95.5 | 66.4 | 62.8 | 87.5 | 77.3 | 70.8 | — | |
| BaseModel=Granite-4.1-8B2026.06 | 73.5 | 93.5 | 68.2 | 63.4 | 87 | 74 | 55.2 | — | |
| PROVEModel=Qwen2.5-7B2026.06 | 73.4 | 81.8 | 60.4 | 63.1 | 88 | 69.7 | 77.1 | — | |
| Qwen2.5-7B + PROVETraining Steps=350 GRPO steps2026.06 | 73.4 | — | — | — | — | — | — | 6.5 | |
| BaseModel=Qwen2.5-7B2026.06 | 66.9 | 73.4 | 54.1 | 57.7 | 79.2 | 62.2 | 75 | — |