| Benchmark | Model | Rank | Results |
|---|---|---|---|
| agents-on-bench | GRPO + ARC (Qwen3-8B) | – | Accuracy: 38.22 |
| agents-on-bench-telecom | GRPO+ARC (Qwen3-4B) | – | Accuracy: 30.03 |
| agents-on-taubench-airline | GRPO + ARC (Qwen3-8B) | – | Accuracy: 44.0 |
| agents-on-taubench-retail | GRPO + ARC (Qwen3-8B) | – | Accuracy: 50.0 |
| instruction-following-on-ifbench | DAPO + ARC (Qwen3-8B) | – | Accuracy: 17.35 |
| mathematical-reasoning-on-aime-2026 | DAPO + ARC (Qwen3-8B) | – | Accuracy: 41.25 |