Proof-pile
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Proof-pile
2.33Perplexity
130
Proof-pile (test)
2.91Perplexity (2K)
16
Proof-pile 128 32k-length documents (test)
2.28Perplexity
8
Proof-Pile
2.915PPL
6
Proof-Pile-2 (PP2)
7.2Tie Percentage
2