Claim Verification on LLMAggreFact (test)
78.1Binary AccuracyThinknCheck
Evaluation results
| Method | Links | |
|---|---|---|
| ThinknCheckModel Size=1B, Precision=fp42026.04 | 78.1 | |
| MiniCheckModel Size=7B, Precision=fp162026.04 | 77.4 | |
| Claude-Sonnet-3.5Evaluation Protocol=zero-shot2026.04 | 77.2 | |
| GPT-4oEvaluation Protocol=zero-shot2026.04 | 75.9 | |
| GPT-4Evaluation Protocol=zero-shot2026.04 | 75.3 | |
| AlignScoreModel Size=355M, Precision=fp162026.04 | 70.4 | |
| ThinknCheck-nothinkModel Size=1B, Precision=fp42026.04 | 57.5 | |
| Gemma3Model Size=1B, Precision=fp4, Evaluation Protocol=base2026.04 | 55.7 | |
| Gemma3 + CoTModel Size=1B, Precision=fp4, Evaluation Protocol=CoT2026.04 | 51.4 |