Fine-grained Hallucination Detection on InFi-Check-FG (test)
92.34BAcc (Normalized)InFi-Checker-Qwen
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| InFi-Checker-Qwenbackbone=Qwen2026.01 | 92.34 | 87.74 | 94.92 | |
| InFi-Checker-Llamabackbone=Llama2026.01 | 90.91 | 85.49 | 94.02 | |
| GPT-4oone-shot=true2026.01 | 64.69 | 60.27 | 80.19 | |
| o3one-shot=true, contains zero-accuracy types=true2026.01 | 60.77 | 56.76 | 94.6 | |
| DeepSeek-V3.2-NoThinkone-shot=true2026.01 | 59.19 | 52.31 | 87.33 | |
| GPT-4.1one-shot=true2026.01 | 58.91 | 55.13 | 94.87 | |
| Qwen3-8Bone-shot=true, contains zero-accuracy types=true2026.01 | 50.49 | 44.78 | 87.22 | |
| Claude-3.7-Sonnetone-shot=true2026.01 | 49.19 | 45.12 | 94.28 | |
| Qwen3-8Bone-shot=false, contains zero-accuracy types=true2026.01 | 47.94 | 41.08 | 86.5 | |
| GPT-5one-shot=true2026.01 | 46.35 | 42.28 | 92.58 | |
| Llama-3.1-8B-Instructone-shot=false, contains zero-accuracy types=true2026.01 | 27.69 | 21.33 | 72.57 | |
| Llama-3.1-8B-Instructone-shot=true, contains zero-accuracy types=true2026.01 | 23.79 | 18.12 | 79.41 | |
| InFi-Checker-Qwenbackbone=Qwen2026.01 | 0.9234 | 0.8774 | 0.9492 | |
| InFi-Checker-Llamabackbone=Llama2026.01 | 0.9091 | 0.8549 | 0.9402 | |
| GPT-4oone-shot settings=true2026.01 | 0.6469 | 0.6027 | 0.8019 | |
| o3one-shot settings=true, contains types with zero accuracy=true2026.01 | 0.6077 | 0.5676 | 0.946 | |
| GPT-4oone-shot settings=false2026.01 | 0.605 | 0.5657 | 0.9497 | |
| DeepSeek-V3.2-NoThinkone-shot settings=false2026.01 | 0.5964 | 0.5394 | 0.8943 | |
| DeepSeek-V3.2-NoThinkone-shot settings=true2026.01 | 0.5919 | 0.5231 | 0.8733 | |
| GPT-4.1one-shot settings=true2026.01 | 0.5891 | 0.5513 | 0.9487 | |
| o3one-shot settings=false, contains types with zero accuracy=true2026.01 | 0.58 | 0.5507 | 0.9548 | |
| GPT-4.1one-shot settings=false2026.01 | 0.5445 | 0.4944 | 0.7817 | |
| Claude-3.7-Sonnetone-shot settings=false2026.01 | 0.5307 | 0.4848 | 0.9221 | |
| Qwen3-8Bone-shot settings=true, contains types with zero accuracy=true2026.01 | 0.5049 | 0.4478 | 0.8722 | |
| Claude-3.7-Sonnetone-shot settings=true2026.01 | 0.4919 | 0.4512 | 0.9428 | |
| GPT-5one-shot settings=false2026.01 | 0.4906 | 0.4441 | 0.9205 | |
| Qwen3-8Bone-shot settings=false, contains types with zero accuracy=true2026.01 | 0.4794 | 0.4108 | 0.865 | |
| GPT-5one-shot settings=true2026.01 | 0.4635 | 0.4228 | 0.9258 | |
| Llama-3.1-8B-Instructone-shot settings=false, contains types with zero accuracy=true2026.01 | 0.2769 | 0.2133 | 0.7257 | |
| Llama-3.1-8B-Instructone-shot settings=true, contains types with zero accuracy=true2026.01 | 0.2379 | 0.1812 | 0.7941 |