HH-SAE factorizes manifolds into nested contextual (L0), atomic (f1), and compository (f2) tiers, achieving 0.9156 cross-domain zero-shot AUC in fraud detection and +9.9% AUPRC lift in steered synthesis.
Incorporating hierarchical semantics in sparse autoencoder architectures.arXiv preprint arXiv:2506.01197
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
fields
cs.LG 6years
2026 6roles
background 2polarities
background 2representative 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.
SFT supplies entangled compositional traces of atomic skills and routing modules; RL identifies those modules and enables recombination on novel compositions outside the SFT support.
Sparse autoencoders provide a basis for sensible concept hierarchies on visual data but are undermined by hard and soft feature absorption.
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.
DACO curates a 15,000-concept dictionary from 400K image-caption pairs and uses it to initialize an SAE that enables granular, concept-specific steering of MLLM activations, raising safety scores on MM-SafetyBench and JailBreakV while preserving general capabilities.
citing papers explorer
-
HH-SAE: Discovering and Steering Hierarchical Knowledge of Complex Manifolds
HH-SAE factorizes manifolds into nested contextual (L0), atomic (f1), and compository (f2) tiers, achieving 0.9156 cross-domain zero-shot AUC in fraud detection and +9.9% AUPRC lift in steered synthesis.
-
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 Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning
SFT supplies entangled compositional traces of atomic skills and routing modules; RL identifies those modules and enables recombination on novel compositions outside the SFT support.
-
Do Sparse Autoencoders Learn Meaningful Concept Hierarchies?
Sparse autoencoders provide a basis for sensible concept hierarchies on visual data but are undermined by hard and soft feature absorption.
-
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.
-
Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs
DACO curates a 15,000-concept dictionary from 400K image-caption pairs and uses it to initialize an SAE that enables granular, concept-specific steering of MLLM activations, raising safety scores on MM-SafetyBench and JailBreakV while preserving general capabilities.