Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Coherence evaluation on HarmBench (test)

0.22N-gram Repetition Rate

DirAbl

-3.692822.718649.1375.5414Jan 27, 2026
Updated 4d ago

Evaluation Results

MethodLinks
2026.01
0.2210053.25
2026.01
1.3899.9947.21
2026.01
1.5899.9546.95
2026.01
1.8299.9969.73
2026.01
2.1999.9646.21
2026.01
3.2110055.88
2026.01
4.1799.9851.31
2026.01
5.0799.9952.78
2026.01
7.2510042.74
2026.01
7.5910041.15
2026.01
13.7799.9139.48
2026.01
18.4930.9321.92
2026.01
19.8310039.67
2026.01
46.2399.9823.3
2026.01
97.071007.53
2026.01
98.041003.2