BatchTopK sparse autoencoders replace per-sample top-k selection with batch-level selection, improving reconstruction at the same average sparsity on GPT-2 and Gemma 2.
Improving dictionary learning with gated sparse autoen- coders, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
BatchTopK Sparse Autoencoders
BatchTopK sparse autoencoders replace per-sample top-k selection with batch-level selection, improving reconstruction at the same average sparsity on GPT-2 and Gemma 2.