Reasoning accuracy on PRONTOQA 3-hop
87AccuracyAVI
Evaluation Results
| Method | Links | |
|---|---|---|
| AVIModel=Qwen 8B2025.05 | 87 | |
| AVIModel=Qwen 4B2025.05 | 68 | |
| No CorrectionModel=Qwen 4B2025.05 | 60 | |
| No CorrectionModel=Qwen 8B2025.05 | 54 | |
| Self CorrectionModel=Qwen 4B2025.05 | 54 | |
| Self CorrectionModel=Qwen 8B2025.05 | 54 |