No article text was recorded for this version.
top-k activation selection
Version 1 of 124 papers27 conceptsfrozen 2026-10-04
Papers in this version
- Blog: Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning
- Blog: k-Sparse Autoencoders
- Blog: Analyzing (In)Abilities of SAEs via Formal Languages
- Blog: Steering CLIP's vision transformer with sparse autoencoders
- Blog: Unpacking SDXL Turbo: Interpreting Text-to-Image Models with Sparse Autoencoders
- Blog: The Lazy Neuron Phenomenon: On Emergence of Activation Sparsity in Transformers
- Blog: Efficient Dictionary Learning with Switch Sparse Autoencoders
- Blog: One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models
- Blog: Winner-Take-All Autoencoders
- Blog: Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation
- Blog: Concept Removal for Frontier Image Generative Models
- Blog: AbsTopK: Rethinking Sparse Autoencoders For Bidirectional Features
- Blog: Residual Stream Analysis with Multi-Layer SAEs
- Blog: Transformer Key-Value Memories Are Nearly as Interpretable as Sparse Autoencoders
- Blog: Group Equivariance Meets Mechanistic Interpretability: Equivariant Sparse Autoencoders
- Blog: Scaling and evaluating sparse autoencoders
- Blog: Learning Deep &#8467;<sub>0</sub> Encoders
- Blog: VFUSE: Virulent Feature Understanding with Sparse autoEncoders
- Blog: Uncovering Branch-specialization in InceptionV1 using k sparse autoencoders
- Blog: Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders
- Blog: Cardinality Restricted Boltzmann Machines
- Blog: Robustly identifying concepts introduced during chat fine-tuning using crosscoders
- Blog: Ultra-High Dimensional Sparse Representations with Binarization for Efficient Text Retrieval
- Blog: Improved modeling of RNA-binding protein motifs in an interpretable neural model of RNA splicing
History & citation
Current · Permanent link
- This direction split from Top-k Sparsity for Mechanistic Interpretability (88% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_top_k_activation_selection_5cba45e6,
title = {top-k activation selection},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/top-k-activation-selection/rdr_5cba45e6e0c64015be64569aa1076a5b},
note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}This exact version (version 1, frozen text):
@misc{lacuna_top_k_activation_selection_5cba45e6_v1,
title = {top-k activation selection},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/top-k-activation-selection/rdr_5cba45e6e0c64015be64569aa1076a5b/v/1},
note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}