Cover Image

Ablating Archetypes: The Stability of Archetypal SAEs is an Artifact of Initialization and Metric Design

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Brzozowski, Michał, Chung, Neo Christopher
Format:	Preprint
Published:	2026
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Aligned Training: A Parameter-Free Method to Improve Feature Quality and Stability of Sparse Autoencoders (SAE)
by: Brzozowski, Michał, et al.
Published: (2026)

Archetypal Analysis++: Rethinking the Initialization Strategy
by: Mair, Sebastian, et al.
Published: (2023)

Reading the Finetuning Prior: Verbatim Content Recovery via Contrastive Decoding Diffing
by: Brzozowski, Michał, et al.
Published: (2026)

GPart: End-to-End Isometric Fine-Tuning via Global Parameter Partitioning
by: Mandica, Paolo, et al.
Published: (2026)

A Survey on Archetypal Analysis
by: Alcacer, Aleix, et al.
Published: (2025)

Archetypal Analysis for Binary Data
by: Wedenborg, A. Emilie J., et al.
Published: (2025)

Incorporating Fairness Constraints into Archetypal Analysis
by: Alcacer, Aleix, et al.
Published: (2025)

Modeling Human Responses by Ordinal Archetypal Analysis
by: Wedenborg, Anna Emilie J., et al.
Published: (2024)

Sparse NMF with Archetypal Regularization: Computational and Robustness Properties
by: Behdin, Kayhan, et al.
Published: (2021)

Archetypal cases for questionnaires with nominal multiple choice questions
by: Alcacer, Aleix, et al.
Published: (2026)

Level the Level: Balancing Game Levels for Asymmetric Player Archetypes With Reinforcement Learning
by: Rupp, Florian, et al.
Published: (2025)

Archetypal Graph Generative Models: Explainable and Identifiable Communities via Anchor-Dominant Convex Hulls
by: Nakis, Nikolaos, et al.
Published: (2026)

Lattice: A Confidence-Gated Hybrid System for Uncertainty-Aware Sequential Prediction with Behavioral Archetypes
by: Bannis, Lorian
Published: (2026)

Discovering EV Charging Site Archetypes Through Few Shot Forecasting: The First U.S.-Wide Study
by: Nikhal, Kshitij, et al.
Published: (2025)

Unveiling Gamer Archetypes through Multi modal feature Correlations and Unsupervised Learning
by: Kanwal, Moona, et al.
Published: (2025)

OptSkills: Learning Generalizable Optimization Skills from Problem Archetypes via Cluster-Based Distillation
by: Yang, Haochen, et al.
Published: (2026)

Robust Player-Conditional Champion Ranking for League of Legends: Style Similarity, Mastery Priors, and Archetype-Constrained Discovery
by: Heo, Min, et al.
Published: (2026)

Riemannian Archetypal Analysis: Interpretable non-linear data analysis on deformed star distributions
by: Diepeveen, Willem, et al.
Published: (2026)

Structural Pass Analysis in Football: Learning Pass Archetypes and Tactical Impact from Spatio-Temporal Tracking Data
by: Karakuş, Oktay, et al.
Published: (2026)

The Rate-Distortion-Polysemanticity Tradeoff in SAEs
by: Mencattini, Tommaso, et al.
Published: (2026)

Tokenized SAEs: Disentangling SAE Reconstructions
by: Dooms, Thomas, et al.
Published: (2025)

Distribution-Aware Feature Selection for SAEs
by: Oozeer, Narmeen, et al.
Published: (2025)

Analyzing (In)Abilities of SAEs via Formal Languages
by: Menon, Abhinav, et al.
Published: (2024)

Regularizing Attention Scores with Bootstrapping
by: Chung, Neo Christopher, et al.
Published: (2026)

Self-Ablating Transformers: More Interpretability, Less Sparsity
by: Ferrao, Jeremias, et al.
Published: (2025)

Mother Archetype
by: Ganganmale, Pramod Akaram
Published: (2026)

Residual Stream Analysis with Multi-Layer SAEs
by: Lawson, Tim, et al.
Published: (2024)

Bayesian algorithmic perfumery: A Hierarchical Relevance Vector Machine for the Estimation of Personalized Fragrance Preferences based on Three Sensory Layers and Jungian Personality Archetypes
by: Martinez, Rolando Gonzales
Published: (2024)

Representation-based Broad Hallucination Detectors Fail to Generalize Out of Distribution
by: Dubanowska, Zuzanna, et al.
Published: (2025)

Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?
by: Korznikov, Anton, et al.
Published: (2026)

Class-Discriminative Attention Maps for Vision Transformers
by: Brocki, Lennart, et al.
Published: (2023)

Resa: Transparent Reasoning Models via SAEs
by: Wang, Shangshang, et al.
Published: (2025)

Can SAEs reveal and mitigate racial biases of LLMs in healthcare?
by: Ahsan, Hiba, et al.
Published: (2025)

Investigating Sensitive Directions in GPT-2: An Improved Baseline and Comparative Analysis of SAEs
by: Lee, Daniel J., et al.
Published: (2024)

Features Emerge as Discrete States: The First Application of SAEs to 3D Representations
by: Miao, Albert, et al.
Published: (2025)

TumorArchetypeR
by: Lütge, Mechthild, et al.
Published: (2026)

SAEs Are Good for Steering -- If You Select the Right Features
by: Arad, Dana, et al.
Published: (2025)

Teach Old SAEs New Domain Tricks with Boosting
by: Koriagin, Nikita, et al.
Published: (2025)

What Sort of a Thing is an Archetype? Archetypes, Complexes and Self‐Organization Revisited
by: Patricia Skar
Published: (2025)

Ablate and Rescue: A Causal Analysis of Residual Stream Hyper-Connections
by: Peng, William, et al.
Published: (2026)