Block-sparse featurizers recover visual concepts as two- to four-dimensional manifolds and describe activations more compactly than direction-based methods via minimum-description-length comparison.
Projecting assumptions: The duality between sparse autoencoders and concept geometry
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 11roles
background 1polarities
background 1representative citing papers
Linear probes for Othello board states factor into tensor-product structure with square and color embeddings composed by a binding matrix, from which the linear probes can be directly recovered.
The paper introduces compositional interpretability as a category-theoretic framework that casts mechanistic explanations as commuting syntactic-semantic mappings optimized under faithfulness and complexity constraints derived from minimum description length.
Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
The paper proposes information scope as a new interpretability axis for SAE features in CLIP and introduces the Contextual Dependency Score to separate local from global scope features, showing they influence model predictions differently.
SAE-NOs extend sparse autoencoders to function spaces via Fourier neural operators with concept and domain sparsity, learning localized patterns more efficiently and generalizing across discretizations on vision data.
Formalizes concept learning in sparse autoencoders as set alignment between human-defined and model-induced concepts, distinguishing detection, separation, and approximation with geometric conditions for neuron representation.
SAEs exhibit a rate-distortion-polysemanticity tradeoff where monosemanticity increases rate and distortion, with optimal polysemanticity set by feature co-occurrence probabilities in the data.
GeoSAE extracts a compact, interpretable feature set from frozen brain MRI foundation models that predicts MCI-to-AD conversion (AUC 0.746) with age-deconfounded annotations and replicates across cohorts.
A new pipeline uses interpretability to characterize concepts in preference data and shape rewards via feature or data interventions during LM post-training.
ViSAE supplies a 64K-image probing suite with 16K concepts, top-down/bottom-up circuit algorithms, and editing methods that raise WaterBirds worst-group accuracy by 48.2% over baselines.
citing papers explorer
-
Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds
Block-sparse featurizers recover visual concepts as two- to four-dimensional manifolds and describe activations more compactly than direction-based methods via minimum-description-length comparison.
-
Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions
Linear probes for Othello board states factor into tensor-product structure with square and color embeddings composed by a binding matrix, from which the linear probes can be directly recovered.
-
From Mechanistic to Compositional Interpretability
The paper introduces compositional interpretability as a category-theoretic framework that casts mechanistic explanations as commuting syntactic-semantic mappings optimized under faithfulness and complexity constraints derived from minimum description length.
-
Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior
Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
-
Beyond Semantics: Disentangling Information Scope in Sparse Autoencoders for CLIP
The paper proposes information scope as a new interpretability axis for SAE features in CLIP and introduces the Contextual Dependency Score to separate local from global scope features, showing they influence model predictions differently.
-
Mechanistic Interpretability with Sparse Autoencoder Neural Operators
SAE-NOs extend sparse autoencoders to function spaces via Fourier neural operators with concept and domain sparsity, learning localized patterns more efficiently and generalizing across discretizations on vision data.
-
A Geometric View for Understanding Concept Learning and Neuron Interpretation in Sparse Autoencoders
Formalizes concept learning in sparse autoencoders as set alignment between human-defined and model-induced concepts, distinguishing detection, separation, and approximation with geometric conditions for neuron representation.
-
The Rate-Distortion-Polysemanticity Tradeoff in SAEs
SAEs exhibit a rate-distortion-polysemanticity tradeoff where monosemanticity increases rate and distortion, with optimal polysemanticity set by feature co-occurrence probabilities in the data.
-
GeoSAE: Geometric Prior-Guided Layer-Wise Sparse Autoencoder Annotation of Brain MRI Foundation Models
GeoSAE extracts a compact, interpretable feature set from frozen brain MRI foundation models that predicts MCI-to-AD conversion (AUC 0.746) with age-deconfounded annotations and replicates across cohorts.
-
Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal
A new pipeline uses interpretability to characterize concepts in preference data and shape rewards via feature or data interventions during LM post-training.
-
Inside the Visual Mind: Neuroscience-Motivated Concept Circuits for Interpreting and Steering Vision Transformers
ViSAE supplies a 64K-image probing suite with 16K concepts, top-down/bottom-up circuit algorithms, and editing methods that raise WaterBirds worst-group accuracy by 48.2% over baselines.