Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Open-ended Text Generation on Wikitext (test)
Loading...
95
Diversity (DIV)
Typical Sampling
-1.72
23.39
48.5
73.61
Oct 27, 2022
Diversity (DIV)
MAUVE Score
Coherence (COH)
Updated 4d ago
Evaluation Results
Method
Method
Links
Diversity (DIV)
MAUVE Score
Coherence (COH)
Typical Sampling
Model=GPT2-XL, Decodin...
2022.10
95
84
53
Nucleus Sampling
Model=OPT-13B, Decodin...
2022.10
92
89
55
Nucleus Sampling
Model=GPT2-XL, Decodin...
2022.10
92
87
57
Contrastive Decoding
Model=OPT-13B, Decodin...
2022.10
91
91
69
Typical Sampling
Model=OPT-13B, Decodin...
2022.10
89
86
58
Contrastive Decoding
Model=GPT2-XL, Decodin...
2022.10
89
92
69
Contrastive Search
Model=OPT-13B, Decodin...
2022.10
87
77
52
Top-k Sampling
Model=GPT2-XL, Decodin...
2022.10
87
79
61
Contrastive Search
Model=GPT2-XL, Decodin...
2022.10
86
75
59
Top-k Sampling
Model=OPT-13B, Decodin...
2022.10
72
77
64
Greedy Decoding
Model=OPT-13B, Decodin...
2022.10
3
8
63
Greedy Decoding
Model=GPT2-XL, Decodin...
2022.10
2
5
62
Feedback
Search any
task
Search any
task