Agent Performance on VitaBench Delivery
29Avg@4NoisyAgent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NoisyAgentbackbone=Qwen3-32B2026.05 | 29 | 43 | |
| DAPObackbone=Qwen3-32B2026.05 | 27.75 | 41 | |
| GSPObackbone=Qwen3-32B2026.05 | 27.25 | 40 | |
| GRPObackbone=Qwen3-32B2026.05 | 27 | 40 | |
| Qwen3-32Bbackbone=Qwen3-32B2026.05 | 23 | 35 | |
| NoisyAgentbackbone=Qwen3-8B2026.05 | 22.25 | 35 | |
| DAPObackbone=Qwen3-8B2026.05 | 21.5 | 34 | |
| GSPObackbone=Qwen3-8B2026.05 | 21.25 | 33 | |
| GRPObackbone=Qwen3-8B2026.05 | 21 | 33 | |
| Qwen3-8Bbackbone=Qwen3-8B2026.05 | 13.75 | 22 |