Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Fact-claim Decomposition on WikiSplitBench-1000 1.0 (test)
Loading...
90.4
Semantic F1
CREDENCE-base
28.936
44.893
60.85
76.807
Jun 18, 2026
Semantic F1
Updated 1mo ago
Evaluation Results
Method
Method
Links
Semantic F1
CREDENCE-base
Model=Qwen3-8B
2026.06
90.4
CREDENCE-all
Model=Qwen3-8B
2026.06
90.3
CREDENCE-base
Model=Gemma-3-12b
2026.06
89.4
CREDENCE-all
Model=Gemma-3-12b
2026.06
89.4
CREDENCE-base
Model=Phi-3-mini
2026.06
89.1
CREDENCE-all
Model=Phi-3-mini
2026.06
89.1
FActScore-prompt
Model=Qwen3-8B
2026.06
87.1
CREDENCE-base
Model=Gemini Flash
2026.06
86.9
CREDENCE-all
Model=Gemini Flash
2026.06
86.9
FActScore-prompt
Model=Gemini Flash
2026.06
85.3
FActScore-prompt
Model=Gemma-3-12b
2026.06
83.6
SpaCy-senter
Model=Phi-3-mini
2026.06
80.9
SpaCy-senter
Model=Qwen3-8B
2026.06
80.9
SpaCy-senter
Model=Gemma-3-12b
2026.06
80.9
SpaCy-senter
Model=Gemini Flash
2026.06
80.9
FActScore-prompt
Model=Phi-3-mini
2026.06
31.3
Feedback
Search any
task
Search any
task