Tool-augmented Question Answering on 2Wiki
51.83AccuracyEPC-AW
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EPC-AWBackbone=Qwen3-Coder-30B2026.05 | 51.83 | — | |
| RollbackBackbone=Qwen3-Coder-30B2026.05 | 44.83 | 7 | |
| RetryBackbone=Qwen3-Coder-30B2026.05 | 42.83 | 9 | |
| No-Repair (AgentFlow)Backbone=Qwen3-Coder-30B2026.05 | 39 | 12.83 |