Knowledge gap detection on TQA
83.2AccuracyGRADEpre
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRADEpreBackbone=Gemma2-9b2026.04 | 83.2 | 93.2 | |
| GRADEpreBackbone=Llama3.1-8b2026.04 | 81.3 | 89.5 | |
| GRADEpreBackbone=Qwen2.5-7b2026.04 | 79.5 | 86.2 | |
| Align-PBackbone=Llama3.1-8b2026.04 | 78.4 | 85.5 | |
| GRADEposBackbone=Gemma2-9b2026.04 | 78.1 | 81.8 | |
| Align-PBackbone=Gemma2-9b2026.04 | 72.5 | 79.9 | |
| GRADEposBackbone=Qwen2.5-7b2026.04 | 71.9 | 77.6 | |
| GRADEposBackbone=Llama3.1-8b2026.04 | 71.3 | 77.4 | |
| Align-PBackbone=Qwen2.5-7b2026.04 | 70.5 | 83.3 | |
| JudgeBackbone=Gemma2-9b2026.04 | 69.2 | — | |
| ICBackbone=Llama3.1-8b2026.04 | 66.1 | 71.5 | |
| JudgeBackbone=Llama3.1-8b2026.04 | 65.2 | — | |
| ICBackbone=Gemma2-9b2026.04 | 64.9 | 69.8 | |
| P-EntropyBackbone=Llama3.1-8b2026.04 | 64.5 | 69.1 | |
| P-EntropyBackbone=Gemma2-9b2026.04 | 63.1 | 65.5 | |
| P-EntropyBackbone=Qwen2.5-7b2026.04 | 61.1 | 64.9 | |
| ICBackbone=Qwen2.5-7b2026.04 | 60.7 | 69.3 | |
| JudgeBackbone=Qwen2.5-7b2026.04 | 60.1 | — |