framework
active
framework:matryoshka-batchtopk-saeMatryoshka BatchTopK SAE
Hierarchical architecture that excels on disentanglement and scales positively.
Neighborhood — ranked by edge-count
Papers (1)
paper
Thinkers (1)
thinker
- Bart BussmannstudiesCo-author of VPD paper; contributes to Goodfire mechanistic interpretability efforts.
Frameworks (2)
framework
- BatchTopK SAEextendsrelated_toEvaluated architecture.
- Matryoshka SAEsrelated_toNested dictionary approach that inspires the nested loss.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Architecture compared.
- Central result.
- Evaluated architecture (Gao et al.).
- Hybrid SAE jointly training turn-averaged inner and per-token outer partitions.
- Sparsity activation used with k=200.
- Sparse dictionary learning method used to extract interpretable features from EEG transformer embeddings.
- Evaluated architecture.
- Interpretable features extracted by sparse autoencoders used as steering targets in this study