Fact-claim Decomposition on SocialClaimSplit-100 1.0 (test)
99.3Semantic F1CREDENCE-base
Evaluation Results
| Method | Links | |
|---|---|---|
| CREDENCE-baseModel=Gemini Flash2026.06 | 99.3 | |
| CREDENCE-allModel=Gemini Flash2026.06 | 99.2 | |
| FActScore-promptModel=Gemma-3-12b2026.06 | 98.2 | |
| FActScore-promptModel=Gemini Flash2026.06 | 97.5 | |
| CREDENCE-allModel=Qwen3-8B2026.06 | 97.3 | |
| FActScore-promptModel=Qwen3-8B2026.06 | 96.9 | |
| CREDENCE-allModel=Gemma-3-12b2026.06 | 95.9 | |
| CREDENCE-baseModel=Gemma-3-12b2026.06 | 95.7 | |
| CREDENCE-baseModel=Qwen3-8B2026.06 | 95.6 | |
| CREDENCE-baseModel=Phi-3-mini2026.06 | 93 | |
| CREDENCE-allModel=Phi-3-mini2026.06 | 92.2 | |
| FActScore-promptModel=Phi-3-mini2026.06 | 87.1 | |
| SpaCy-senterModel=Phi-3-mini2026.06 | 81.1 | |
| SpaCy-senterModel=Qwen3-8B2026.06 | 81.1 | |
| SpaCy-senterModel=Gemma-3-12b2026.06 | 81.1 | |
| SpaCy-senterModel=Gemini Flash2026.06 | 81.1 |