What Do Sparse Autoencoder Feature Discovery Scores Identify? From Activations to Interventions
cs.LG, cs.CL, stat.ML
Submitted: 2026-10-11
Updated: 2026-10-11
Sources
- Disentangled Feature Importance
- Which Sparse Autoencoder Features Are Real? Model-X Knockoffs for False Discovery Rate Control
- How to use and interpret activation patching
- AtP*: An efficient and scalable method for localizing LLM behaviour to components
- Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders
- Mechanistic Interpretability of Code Correctness in LLMs via Sparse Autoencoders
- Decodability is Not Causality: Dissociating Probe Readouts from Behavioral Drivers via SAE Decomposition
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks