Domain-Filtered Knowledge Graphs from Sparse Autoencoder Features
Sparse autoencoders (SAEs) extract millions of interpretable features from a language model, but flat feature inventories aren't very useful on their own. Domain concepts get mixed with generic and weakly grounded features, while related ideas are scattered across many units, and there's no way to understand relationships between features. We address this by first constructing a strict domain-specific concept universe from a large SAE inventory using contrastive activations and a multi-stage filtering process. Next, we build two aligned graph views on the filtered set: a co-occurrence graph for corpus-level conceptual structure, organized at multiple levels of granularity, and a transcoder-based mechanism graph that links source-layer and target-layer features through sparse latent pathways. Automated edge labeling then turns these graph views into readable knowledge graphs rather than unlabeled layouts. In a case study on a biology textbook, these graphs recover coherent chapter and subchapter-level structure, reveal concepts that bridge neighboring topics, and transform messy sentence-level activity containing thousands of features into compact, readable views that illustrate the model's local activity. Taken together, this reframes a flat SAE inventory as an internal knowledge graph that converts feature-level interpretability into a global map of model knowledge and enables audits of reasoning faithfulness.
Code (0)
등록된 구현이 없습니다.
Tasks
Knowledge GraphsSimilar Papers 제목 키워드 기반
Causal Concept Graphs in LLM Latent Space for Stepwise Reasoning
Sparse autoencoders can localize where concepts live in language models, but not how they interact during multi-step reasoning. We propose Causal Concept Graphs (CCG): a directed acyclic graph over sparse, interpretable …
MMPKUBase: A Comprehensive and High-quality Chinese Multi-modal Knowledge Graph
Multi-modal knowledge graphs have emerged as a powerful approach for information representation, combining data from different modalities such as text, images, and videos. While several such graphs have been constructed …
AttributeContrastive LearningKnowledge GraphsMulti-modal Knowledge Graph+3Efficient Variational Graph Autoencoders for Unsupervised Cross-domain Prerequisite Chains
Prerequisite chain learning helps people acquire new knowledge efficiently. While people may quickly determine learning paths over concepts in a domain, finding such paths in other domains can be challenging. We introduc…
Link PredictionBlind Image Deconvolution using Student's-t Prior with Overlapping Group Sparsity
In this paper, we solve blind image deconvolution problem that is to remove blurs form a signal degraded image without any knowledge of the blur kernel. Since the problem is ill-posed, an image prior plays a significant …
Image DeconvolutionDropout Training of Matrix Factorization and Autoencoder for Link Prediction in Sparse Graphs
Matrix factorization (MF) and Autoencoder (AE) are among the most successful approaches of unsupervised learning. While MF based models have been extensively exploited in the graph modeling and link prediction literature…
Link PredictionMultiview Learning