Binary Factuality Verification on TriviaQA
88.58AccuracyDiVA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiVAModel=Qwen2.5-7B-Inst2026.01 | 88.58 | 88.53 | |
| DiVAModel=Qwen3-4B-Inst2026.01 | 88.19 | 88.17 | |
| AG-VerifierModel=Qwen2.5-7B-Inst2026.01 | 75.98 | 74.89 | |
| AG-VerifierModel=Qwen3-4B-Inst2026.01 | 74.41 | 73.66 | |
| G-VerifierModel=Qwen3-4B-Inst2026.01 | 66.38 | 65.45 | |
| G-VerifierModel=Qwen2.5-7B-Inst2026.01 | 66.14 | 64.82 | |
| FactScoreModel=GPT-42026.01 | 59.95 | 52.85 |