Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Sparse Autoencoder Evaluation on Pythia-160M activations SAEBench downstream performance
Loading...
3.906
CE Loss
iSAE-ME
3.90208
3.92854
3.955
3.98146
May 29, 2026
CE Loss
Sparse Probing Accuracy
SCR
TPP
Updated 2d ago
Evaluation Results
Method
Method
Links
CE Loss
Sparse Probing Accuracy
SCR
TPP
iSAE-ME
Backbone=Pythia-160M,...
2026.05
3.906
90.8
29.3
4.4
TopK
Backbone=Pythia-160M
2026.05
3.974
90.9
38.1
4.3
iSAE
Backbone=Pythia-160M,...
2026.05
4.003
90.7
34.8
4.3
AbsTopK
Backbone=Pythia-160M
2026.05
4.004
90.6
33
3.1
Feedback
Search any
task
Search any
task
