Tool-use on TAU2
31.3ScoreGranite-4.1-8B + PROVE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Granite-4.1-8B + PROVETraining Steps=350 GRPO steps2026.06 | 31.3 | 6.4 | |
| Qwen3-8B + PROVETraining Steps=350 GRPO steps2026.06 | 29.3 | 3.5 | |
| Qwen3-4B + PROVETraining Steps=350 GRPO steps2026.06 | 28.6 | 3.6 | |
| Qwen2.5-7B + PROVETraining Steps=350 GRPO steps2026.06 | 21.8 | 6.8 |