Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Code Generation on HumanEval (ΔPPL)
Loading...
14.8
ΔPPL (%)
CE
-1.424
2.788
7
11.212
May 28, 2026
ΔPPL (%)
Updated 5d ago
Evaluation Results
Method
Method
Links
ΔPPL (%)
CE
Backbone=0.5B, Adapter...
2026.05
14.8
CE
Backbone=0.5B, Adapter...
2026.05
14.3
CE
Backbone=0.5B, Adapter...
2026.05
13.9
CE
Backbone=7B, Adapter=LoRA
2026.05
12.8
CE
Backbone=7B, Adapter=DoRA
2026.05
12.5
CE
Backbone=7B, Adapter=S...
2026.05
12.1
CE
Backbone=0.5B, Adapter...
2026.05
10.5
CE
Backbone=7B, Adapter=R...
2026.05
9.5
Base
Backbone=0.5B
2026.05
5.12
Base
Backbone=7B
2026.05
3.45
CE + TMKL
Backbone=0.5B, Adapter...
2026.05
0.8
CE + TMKL
Backbone=0.5B, Adapter...
2026.05
0.7
CE + TMKL
Backbone=0.5B, Adapter...
2026.05
0.5
CE + TMKL
Backbone=0.5B, Adapter...
2026.05
0.2
CE + TMKL
Backbone=7B, Adapter=R...
2026.05
0.1
CE + TMKL
Backbone=7B, Adapter=S...
2026.05
-0.5
CE + TMKL
Backbone=7B, Adapter=DoRA
2026.05
-0.6
CE + TMKL
Backbone=7B, Adapter=LoRA
2026.05
-0.8
Feedback
Search any
task
Search any
task
