Neuronpedia
Get Started
Releases
Evals
Steer
Search
Privacy & Terms
Contact
Sign In
SAE RELEASE
Sparse Autoencoders for Pythia-70M-Deduped
Pythia-70M-Deduped
· Unnamed
anonymous.4open.science ↗
Jump to Feature
Attention Out
SAE SET
LAYER
INDEX
Go
Search + Test
Attention Out
SAE SET
All Layers
LAYER
SEARCH
Browse SAE Sets
PYTHIA-70M-DEDUPED - Attention Out
(
att-sm
)
Layer 0
0-att-sm
Layer 1
1-att-sm
Layer 2
2-att-sm
Layer 3
3-att-sm
Layer 4
4-att-sm
Layer 5
5-att-sm
PYTHIA-70M-DEDUPED - MLP Post
(
mlp-sm
)
Layer 0
0-mlp-sm
Layer 1
1-mlp-sm
Layer 2
2-mlp-sm
Layer 3
3-mlp-sm
Layer 4
4-mlp-sm
Layer 5
5-mlp-sm
PYTHIA-70M-DEDUPED - Residuals
(
res-sm
)
Layer 0
0-res-sm
Layer 1
1-res-sm
Layer 2
2-res-sm
Layer 3
3-res-sm
Layer 4
4-res-sm
Layer 5
5-res-sm
Layer Embed
e-res-sm
© Neuronpedia 2024
Privacy & Terms
Twitter
Contact