Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Fact Generation (Scratch) on WD50K
Loading...
71.7
Accuracy
KREPE
15.332
29.966
44.6
59.234
May 22, 2026
Accuracy
Updated 8d ago
Evaluation Results
Method
Method
Links
Accuracy
KREPE
LLM=N/A
2026.05
71.7
Neighbor Sets
LLM=Gemini 3 Pro
2026.05
35.1
Few-shot Facts
LLM=Gemini 3 Pro
2026.05
32.6
Random Facts
LLM=Gemini 3 Pro
2026.05
31.7
Re-ranking
LLM=Gemini 3 Pro
2026.05
30.4
Gibbs Sampling
LLM=N/A
2026.05
29.9
Random Facts
LLM=GPT 5.2
2026.05
29.8
Few-shot Facts
LLM=GPT 5.2
2026.05
29
Autoregressive
LLM=Gemini 3 Pro
2026.05
26.5
Iterative Prediction
LLM=N/A
2026.05
26.2
Autoregressive
LLM=GPT 5.2
2026.05
24.8
Re-ranking
LLM=GPT 5.2
2026.05
21.5
Neighbor Sets
LLM=GPT 5.2
2026.05
17.5
Feedback
Search any
task
Search any
task