Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Toxicity Classification on CivilComments Hard subset - demographic identities
Loading...
86.9
Hard Accuracy
Robust Filtered
85.236
85.668
86.1
86.532
Jun 18, 2026
Hard Accuracy
Hard ECE
Hard ECE+TS
Updated 1mo ago
Evaluation Results
Method
Method
Links
Hard Accuracy
Hard ECE
Hard ECE+TS
Robust Filtered
2026.06
86.9
4
3.1
Robust MoE
2026.06
86.8
3.7
2.9
Single Expert
2026.06
86.7
10.5
9.7
Vanilla MoE
2026.06
86.6
10.8
10
MoCaE
2026.06
86.6
10.1
9.1
FGR
2026.06
85.7
6.5
5.3
FGR + Robust
2026.06
85.3
6.5
5.7
Feedback
Search any
task
Search any
task