Agent Performance on VitaBench In-Store
32.25Avg@4NoisyAgent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NoisyAgentbackbone=Qwen3-32B2026.05 | 32.25 | 46 | |
| GSPObackbone=Qwen3-32B2026.05 | 30.5 | 44 | |
| GRPObackbone=Qwen3-32B2026.05 | 30.25 | 43 | |
| DAPObackbone=Qwen3-32B2026.05 | 29.5 | 42 | |
| Qwen3-32Bbackbone=Qwen3-32B2026.05 | 26 | 38 | |
| NoisyAgentbackbone=Qwen3-8B2026.05 | 24 | 37 | |
| DAPObackbone=Qwen3-8B2026.05 | 23.25 | 36 | |
| GSPObackbone=Qwen3-8B2026.05 | 23 | 35 | |
| GRPObackbone=Qwen3-8B2026.05 | 22.75 | 35 | |
| Qwen3-8Bbackbone=Qwen3-8B2026.05 | 15.5 | 24 |