Sparse Autoencoders (SAEs)

Applied to Proof-of-Concept Debugger for a Small LLM by StefanHex 21h ago

Applied to Topological Data Analysis and Mechanistic Interpretability by Jakob Hansen 17d ago

Applied to Takeaways From Our Recent Work on SAE Probing by Josh Engels 17d ago

Applied to SAE Training Dataset Influence in Feature Matching and a Hypothesis on Position Features by Seonglae Cho 23d ago

Applied to [PAPER] Jacobian Sparse Autoencoders: Sparsify Computations, Not Just Activations by Lucy Farnik 23d ago

Applied to Deep sparse autoencoders yield interpretable features too by Armaan A. Abraham 26d ago

Applied to Sparse Autoencoder Features for Classifications and Transferability by Shan23Chen 1mo ago

Applied to Comparing the effectiveness of top-down and bottom-up activation steering for bypassing refusal on harmful prompts by RobertM 1mo ago

Applied to Cross-Layer Feature Alignment and Steering in Large Language Model by dlaptev 1mo ago

Applied to SAE regularization produces more interpretable models by Logan Riggs 2mo ago

Applied to Empirical Insights into Feature Geometry in Sparse Autoencoders by Jason Boxi Zhang 2mo ago

Applied to Finding Features Causally Upstream of Refusal by Daniel Lee 2mo ago

Applied to Scaling Sparse Feature Circuit Finding to Gemma 9B by Diego Caples 2mo ago

Applied to Broken Latents: Studying SAEs and Feature Co-occurrence in Toy Models by chanind 3mo ago

Applied to Are Sparse Autoencoders a good idea for AI control? by Gerard Boxo 3mo ago

Applied to Learning Multi-Level Features with Matryoshka SAEs by Bart Bussmann 3mo ago

Applied to Compositionality and Ambiguity:  Latent Co-occurrence and Interpretable Subspaces by Matthew A. Clarke 3mo ago

Applied to Matryoshka Sparse Autoencoders by Noa Nabeshima 3mo ago

Applied to Measuring Nonlinear Feature Interactions in Sparse Crosscoders [Project Proposal] by Jason Gross 3mo ago