REVIEW 9 cited by
The Missing Curve Detectors of InceptionV1: Applying Sparse Autoencoders to InceptionV1 Early Vision
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Recent work on sparse autoencoders (SAEs) has shown promise in extracting interpretable features from neural networks and addressing challenges with polysemantic neurons caused by superposition. In this paper, we apply SAEs to the early vision layers of InceptionV1, a well-studied convolutional neural network, with a focus on curve detectors. Our results demonstrate that SAEs can uncover new interpretable features not apparent from examining individual neurons, including additional curve detectors that fill in previous gaps. We also find that SAEs can decompose some polysemantic neurons into more monosemantic constituent features. These findings suggest SAEs are a valuable tool for understanding InceptionV1, and convolutional neural networks more generally.
Forward citations
Cited by 9 Pith papers
-
SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoders
A supervised sparse autoencoder binds each concept to a single neuron, letting Stable Diffusion erase a concept by steering one latent.
-
Relevance-driven Input Dropout: an Explanation-guided Regularization Technique
RelDrop, which occludes the most attribution-relevant input regions during training, improves generalization and occlusion robustness for image and point cloud classification.
-
SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders
SAeUron removes concepts from text-to-image diffusion models by ablating concept-specific sparse autoencoder features during inference, achieving state-of-the-art unlearning on UnlearnCanvas and I2P without weight updates.
-
Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models
Dictionary learning with ICFL and control-data whitening extracts sparse features from microscopy foundation models that correlate with cell types and genetic perturbations.
-
Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders
Sparse autoencoders with single-layer encoders provably incur a non-zero amortisation gap, and more expressive encoders (MLPs, iterative optimization) improve both feature recovery and interpretability.
-
InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoders
Sparse autoencoders trained on ESM-2 recover thousands of interpretable features that align with Swiss-Prot concepts and can influence sequence generation.
-
Evaluating SAE interpretability without explanations
SAE latent interpretability can be scored directly from activation examples via intruder detection and embedding clustering, with LLM scores correlating strongly with human scores.
-
Interpreting and Steering Protein Language Models through Sparse Autoencoders
Amplifying specific sparse-autoencoder latents inside ESM-2 biases sequence generation toward zinc finger motifs, with a low success rate and incomplete statistical validation.
-
A Monosemantic Attribution Framework for Stable Interpretability in Clinical Neuroscience Transformer-Based Language Models
Using a sparse autoencoder bottleneck before attribution lowers reported RIS/ROS instability scores for a ModernBERT Alzheimer's classifier, but the improvement is measured with the optimizer's own training objective.
Discussion (0). Continue with ORCID to comment.