Global Fact Consistency Verification on VitaminC
0.984PrecisionDirect (baseline)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Direct (baseline)Backbone=GPT-OSS-120B2026.01 | 0.984 | 0.926 | 0.953 | |
| Direct (baseline)Backbone=DeepSeek-R12026.01 | 0.98 | 0.73 | 0.827 | |
| Direct (baseline)Backbone=Claude 3.72026.01 | 0.979 | 0.854 | 0.909 | |
| QXRBackbone=DeepSeek-R12026.01 | 0.973 | 0.99 | 0.981 | |
| QXRBackbone=Mistral Large (2407)2026.01 | 0.968 | 0.978 | 0.972 | |
| QXRBackbone=Claude 3.72026.01 | 0.956 | 0.975 | 0.965 | |
| Direct (baseline)Backbone=Claude 42026.01 | 0.956 | 0.877 | 0.913 | |
| QXRBackbone=GPT-OSS-120B2026.01 | 0.956 | 0.995 | 0.975 | |
| Direct (baseline)Backbone=Mistral Large (2407)2026.01 | 0.955 | 0.603 | 0.724 | |
| QXRBackbone=Claude 42026.01 | 0.938 | 0.983 | 0.96 |