Tool-augmented Question Answering on Bamboogle (Accuracy, Delta)
58.13AccuracyEPC-AW
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EPC-AWBackbone=Qwen3-Coder-30B2026.05 | 58.13 | — | |
| RollbackBackbone=Qwen3-Coder-30B2026.05 | 52.27 | 5.86 | |
| RetryBackbone=Qwen3-Coder-30B2026.05 | 51.47 | 6.66 | |
| No-Repair (AgentFlow)Backbone=Qwen3-Coder-30B2026.05 | 48.27 | 9.86 |