Single SAEs trained on the residual stream activation vectors from every transformer layer simultaneously: https://arxiv.org/abs/2409.04185
Tim Lawson
tim-lawson
AI & ML interests
Mechanistic interpretability, language modelling, semantics
Organizations
None yet
models
75
tim-lawson/mlsae-pythia-70m-deduped-x64-k256-lens
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k256-lens-tfm
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k128-lens
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k128-lens-tfm
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k64-lens
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k64-lens-tfm
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k16-lens
Updated
tim-lawson/mlsae-pythia-70m-deduped-x64-k16-lens-tfm
Updated
tim-lawson/mlsae-pythia-1b-deduped-x64-k32
Updated
tim-lawson/mlsae-pythia-1b-deduped-x64-k32-tfm
Updated
•
2
datasets
39
tim-lawson/mlsae-pythia-410m-deduped-x64-k32-dists
Viewer
•
Updated
•
65.5k
•
1
tim-lawson/mlsae-pythia-160m-deduped-x64-k32-lens-dists
Viewer
•
Updated
•
49.2k
•
7
tim-lawson/mlsae-pythia-70m-deduped-x64-k32-lens-dists
Viewer
•
Updated
•
32.8k
•
7
tim-lawson/mlsae-pythia-70m-deduped-x64-k256-examples
Viewer
•
Updated
•
3.63M
tim-lawson/mlsae-lens-std-pythia-70m-deduped-x64-k32-dists
Viewer
•
Updated
•
32.8k
•
1
tim-lawson/mlsae-lens-nostd-pythia-70m-deduped-x64-k32-dists
Viewer
•
Updated
•
32.8k
•
1
tim-lawson/mlsae-pythia-160m-deduped-x32-k32-examples
Viewer
•
Updated
•
8.31M
tim-lawson/mlsae-pythia-70m-deduped-x64-k128-examples
Viewer
•
Updated
•
5.78M
tim-lawson/mlsae-pythia-70m-deduped-x64-k64-examples
Viewer
•
Updated
•
5.5M
•
2
tim-lawson/mlsae-pythia-160m-deduped-x8-k32-examples
Viewer
•
Updated
•
2.31M
•
2