mlsanityData to 14 Oct 2026

Domain-Filtered Knowledge Graphs from Sparse Autoencoder Features

arXiv 2604.23829 ↗First posted 26 Apr 2026cs.AIcited by 0

On the map: Mechanistic interpretability → SAE features and their applications

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in SAE features and their applications, one of 13 directions in Mechanistic interpretability. Builds on 19 of its 42 references that are in the map. First posted on arXiv 26 Apr 2026.

Builds on

19 of 42 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability12 Mar
  2. 2024Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoders27 Oct
  3. 2024Evolution of SAE Features Across Layers in LLMs11 Oct
  4. 2024Scaling and evaluating sparse autoencoders6 Jun
  5. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Models15 Sep

Other references · 14

  1. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,122
  2. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 703
  3. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 435
  4. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 394
  5. 2024Transcoders Find Interpretable LLM Feature Circuitscited by 200
  6. 2022PINTO: Faithful Language Reasoning Using Prompt-Generated Rationalescited by 56
  7. 2024The Geometry of Concepts: Sparse Autoencoder Feature Structurecited by 52
  8. 2018A Survey on Open Information Extractioncited by 51
  9. 2025Enhancing Automated Interpretability with Output-Centric Feature Descriptionscited by 41
  10. 2025Analyze Feature Flow to Enhance Interpretation and Steering in Language Modelscited by 13
  11. 2025Model Directions, Not Words: Mechanistic Topic Models Using Sparse Autoencoderscited by 8
  12. 2025LLMs are Overconfident: Evaluating Confidence Interval Calibration with FermiEvalcited by 8
2 more
  1. 2025Demystifying Errors in LLM Reasoning Traces: An Empirical Study of Code Execution Simulationcited by 5
  2. 2026Flash-SD-KDE: Accelerating SD-KDE with Tensor Corescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026From Token Lists to Graph Motifs: Weisfeiler-Lehman Analysis of Sparse Autoencoder Features
  2. 2026When Trees Are Not Enough: Learning Mixed-Topology Feature Graphs with Adaptive Graph Sparse Autoencoders
  3. 2024Disentangling Dense Embeddings with Sparse Autoencoders
  4. 2026From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse Autoencoders
  5. 2026Navigating the Concept Space of Language Models
  6. 2026Causal Concept Graphs in LLM Latent Space for Stepwise Reasoning
  7. 2025Temporal Sparse Autoencoders: Leveraging the Sequential Nature of Language for Interpretability
  8. 2025Disentangling concept semantics via multilingual averaging in Sparse Autoencoders