arXiv Machine Learning July 24, 2026 By Larry Richards

Do Active SAE Feature Planes Carry More Holonomy? A Preregistered Reversal in Gemma

Read the original on arXiv Machine Learning →

arXiv:2607. 20522v1 Announce Type: new Abstract: This paper tests whether holonomy concentrates on active sparse-autoencoder (SAE) feature planes in Gemma 2 2B, a concrete operationalization of the broader semantic-concentration prediction.

Summary generated by The Flow from the publisher's feed. The full article lives at arXiv Machine Learning.

arXiv AI

Jun 9

Pre-Intervention Prediction of Sparse Autoencoder Steering Side Effects

arXiv:2606. 08365v1 Announce Type: cross Abstract: Sparse autoencoder (SAE) features are increasingly used to steer language models, but feature steering is rarely clean: the same intervention can behave inconsistently across contexts and perturb unrelated features.

By Evan Duan

llms

arXiv AI

Jul 14

When Are Sparse Feature Interventions Actually Localized? Matched Evaluation for SAE-Based Safety Control

arXiv:2607. 10226v1 Announce Type: new Abstract: We evaluate when sparse autoencoder (SAE) features act as localized control handles for safety-relevant behavior.

By Daming Luo

benchmarks safety

arXiv Machine Learning

Jul 24

Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects

arXiv:2607. 20596v1 Announce Type: new Abstract: Sparse autoencoder (SAE) features are used to interpret and steer large language models, yet whether a feature's causal role is stable across SAE families remains untested.

By Seonglae Cho, Zekun Wu, Kleyton Da Costa, Rishi Kalra, Ilham Wicaksono, Adriano Koshiyama

llms benchmarks safety

arXiv AI

Jun 18

SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior

arXiv:2606. 18322v1 Announce Type: cross Abstract: Sparse Autoencoders (SAEs) decompose residual-stream activations into interpretable features.

By Mingyue Cui, Linghui Shen, Xingyi Yang

safety

arXiv Machine Learning

Jul 28

Steering grids for sparse-autoencoder features: when a top-context label names an activation regime rather than a causal axis

arXiv:2605. 03160v2 Announce Type: replace Abstract: The standard protocol for interpreting sparse-autoencoder (SAE) features labels each feature from its top-activating contexts and validates the label by steering that single feature at a typical magnitude.

By Michael A. Riegler, Birk Sebastian Frostelid Torpmann-Hagen

llms

arXiv Machine Learning

Aug 4

Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations

arXiv:2605. 28149v2 Announce Type: replace Abstract: Sparse Autoencoders (SAEs) extract interpretable features from Large Language Model activations, but standard variants enforce non-negative latents, so a bidirectional semantic axis (e.

By Bartosz Wieciech, Zmnako Awrahman, Marcin Czelej, Victor Hugo Jaramillo Velasquez, Wioletta Stobieniecka

llms efficiency safety

Related stories

Pre-Intervention Prediction of Sparse Autoencoder Steering Side Effects

When Are Sparse Feature Interventions Actually Localized? Matched Evaluation for SAE-Based Safety Control

Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects

SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior

Steering grids for sparse-autoencoder features: when a top-context label names an activation regime rather than a causal axis

Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations