Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Natural Language Generation on UltraChat
Loading...
36.8
BLEU
ADORE
20.472
24.711
28.95
33.189
Jul 2, 2024
BLEU
ROUGE
BERT-F
Updated 3d ago
Evaluation Results
Method
Method
Links
BLEU
ROUGE
BERT-F
ADORE
Backbone=Llama-2 7B, C...
2024.07
36.8
28.8
63.5
Full Attention
Backbone=Llama-2 7B, C...
2024.07
35.6
29.2
63.4
Strided Attention
Backbone=Llama-2 7B, C...
2024.07
28
24.8
57.5
H2O(Rebuilt)
Backbone=Llama-2 7B, C...
2024.07
27.6
26.9
61.5
Window Attention
Backbone=Llama-2 7B, C...
2024.07
26.7
28
61.4
H2O
Backbone=Llama-2 7B, C...
2024.07
26.4
25.3
60.3
StreamingLLM
Backbone=Llama-2 7B, C...
2024.07
23.9
26
59.6
KV Compression
Backbone=Llama-2 7B, C...
2024.07
21.1
23.2
56.9
Feedback
Search any
task
Search any
task