General QA Verification on PopQA
90.14P@1DiVA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiVABackbone=Qwen-2.5-7B-Instruct, AR=true, EK=true, FS=true, TE=true2026.01 | 90.14 | 85.21 | |
| DiVABackbone=Qwen-3-4B-Instruct, AR=true, EK=true, FS=true, TE=true2026.01 | 88.77 | 83.2 | |
| DiVA w/o compressionBackbone=Qwen-3-4B-Instruct, AR=true, EK=true, FS=true, TE=true2026.01 | 87.67 | 82.1 | |
| DiVA w/o compressionBackbone=Qwen-2.5-7B-Instruct, AR=true, EK=true, FS=true, TE=true2026.01 | 83.56 | 79.91 | |
| Agentic-Generative VerifierBackbone=Qwen-3-4B-Instruct, AR=true, EK=true, FS=false, TE=true2026.01 | 82.47 | 76.99 | |
| Agentic-Generative ScorerBackbone=Qwen-2.5-7B-Instruct, AR=true, EK=true, FS=true, TE=false2026.01 | 81.37 | 73.33 | |
| Agentic-Generative ScorerBackbone=Qwen-3-4B-Instruct, AR=true, EK=true, FS=true, TE=false2026.01 | 81.37 | 73.33 | |
| Agentic-Generative VerifierBackbone=Qwen-2.5-7B-Instruct, AR=true, EK=true, FS=false, TE=true2026.01 | 81.1 | 76.26 | |
| Discriminative VerifierBackbone=Qwen-2.5-7B-Instruct, AR=false, EK=false, FS=true, TE=true2026.01 | 73.42 | 61.1 | |
| Generative VerifierBackbone=Qwen-2.5-7B-Instruct, AR=true, EK=false, FS=false, TE=true2026.01 | 71.15 | 58.42 | |
| FactScore w/ GPT-4Backbone=GPT-42026.01 | 70.14 | 61.32 | |
| Generative VerifierBackbone=Qwen-3-4B-Instruct, AR=true, EK=false, FS=false, TE=true2026.01 | 68.07 | 57.24 | |
| MiniCheck w/ MiniCheck-7BBackbone=MiniCheck-7B2026.01 | 67.4 | 57.44 | |
| Discriminative VerifierBackbone=Qwen-3-4B-Instruct, AR=false, EK=false, FS=true, TE=true2026.01 | 66.85 | 52.51 | |
| HHEM-2.1-Open2026.01 | 57.53 | 45.57 | |
| MiniCheck w/ Flan-T5Backbone=Flan-T52026.01 | 56.99 | 47.4 |