A Universal Vibe? Finding and Controlling Language-Agnostic Informal Register with SAEs
Paper • 2603.26236 • Published
Interpretability & SAEs
-
-
Parts-of-Speech as Emergent Categories in SAE Latent Space
Paper • 2609.29362 • Published • 11 -
PolySAE: Modeling Feature Interactions in Sparse Autoencoders via Polynomial Decoding
Paper • 2602.01322 • Published • 8 -
SAE-V: Interpreting Multimodal Models for Enhanced Alignment
Paper • 2502.17514 • Published -
SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior
Paper • 2606.18322 • Published • 15 -
SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders
Paper • 2501.18052 • Published • 8 -
SAEs Are Good for Steering -- If You Select the Right Features
Paper • 2505.20063 • Published • 1 -
SAEdit: Token-level control for continuous image editing via Sparse AutoEncoder
Paper • 2510.05081 • Published • 5 -
VL-SAE: Interpreting and Enhancing Vision-Language Alignment with a Unified Concept Set
Paper • 2510.21323 • Published • 4 -
Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders
Paper • 2605.07922 • Published -
Rethinking the Diffusion Model from a Langevin Perspective
Paper • 2604.10465 • Published • 16 -
Distribution Matching Variational AutoEncoder
Paper • 2512.07778 • Published • 29 -
Interpretability as Compression: Reconsidering SAE Explanations of Neural Activations with MDL-SAEs
Paper • 2410.11179 • Published • 2 -
Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?
Paper • 2602.14111 • Published • 57