Agent Performance on AgentNoiseBench-Vita Noisy setting 1.0 (test)
28.75Delivery Avg@4 ScoreNoisyAgent
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| NoisyAgentBackbone Model=Qwen3-32B, Training Method=NoisyAgent2026.05 | 28.75 | 42 | 22 | 31 | 9.5 | 14 | |
| DAPOBackbone Model=Qwen3-32B, Training Method=DAPO2026.05 | 24 | 36 | 16.75 | 24 | 7.5 | 11 | |
| GSPOBackbone Model=Qwen3-32B, Training Method=GSPO2026.05 | 23.75 | 36 | 17.5 | 25 | 7.5 | 12 | |
| GRPOBackbone Model=Qwen3-32B, Training Method=GRPO2026.05 | 23.25 | 35 | 19.5 | 28 | 7.25 | 11 | |
| NoisyAgentBackbone Model=Qwen3-8B, Training Method=NoisyAgent2026.05 | 21.5 | 34 | 16.25 | 25 | 4.75 | 8 | |
| Qwen3-32BBackbone Model=Qwen3-32B, Training Method=Base2026.05 | 19.5 | 30 | 14.75 | 21 | 5.5 | 9 | |
| GSPOBackbone Model=Qwen3-8B, Training Method=GSPO2026.05 | 16 | 26 | 15 | 22 | 2.75 | 5 | |
| DAPOBackbone Model=Qwen3-8B, Training Method=DAPO2026.05 | 15.75 | 25 | 12.75 | 19 | 2.25 | 4 | |
| GRPOBackbone Model=Qwen3-8B, Training Method=GRPO2026.05 | 15.25 | 24 | 14.25 | 23 | 2.5 | 4 | |
| Qwen3-8BBackbone Model=Qwen3-8B, Training Method=Base2026.05 | 11.75 | 18 | 8.5 | 12 | 0.75 | 2 |