mlsanityData to 7 Oct 2026

From Input to Output: A Flexible Agent for Dual-End Interpretation of Sparse Autoencoder Features

arXiv 2609.35367 ↗First posted 28 Sep 2026cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Sparse autoencoders (SAEs) are an important tool for mechanistic interpretability, but interpreting their many features remains challenging. Existing methods characterize input-side activation patterns and output-side intervention effects, yet often leave their functional connection implicit, while input-side evidence collection typically relies on costly large-corpus scans. We introduce functional interpretation, which characterizes an SAE feature as a mapping from its activating input semantics to its output effects under intervention, and present Dual-End Agentic Feature Interpretation (DAFI), an agent that actively gathers evidence and refines input-side, output-side, and functional interpretations through component-specific feedback. Its short-context token probing enables on-demand activation evidence collection without a full corpus scan. On GemmaScope, DAFI improves Input score by 13.1 percentage points over SAGE and Output score by 38.9 points over Token Change, while being substantially more token-efficient than a general-purpose coding agent. Skills distilled from successful refinements raise the held-out joint pass rate from 58.0% to 92.0% and improve both interpretation quality and efficiency when transferred to a new model-SAE setting. Across features with reliable endpoint interpretations, 70.7% exhibit non-equivalent input and output semantics. On AxBench, DAFI also improves steering-feature selection over output-score filtering. Code is available at https://github.com/THUAIS-Lab/DAFI.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 28 of its 36 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

28 of 36 references.

References · 28

  1. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  2. 2025Qwen3 Technical Reportcited by 8,195
  3. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  4. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  5. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,928
  6. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  7. 2022Toy Models of Superpositioncited by 928
  8. 2026DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligencecited by 877
  9. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 814
  10. 2024LLM Evaluators Recognize and Favor Their Own Generationscited by 651
  11. 2024Scaling and evaluating sparse autoencoderscited by 530
  12. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
16 more
  1. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 381
  2. 2024Automated Design of Agentic Systemscited by 360
  3. 2025Agentic Context Engineering: Evolving Contexts for Self-Improving Language Modelscited by 349
  4. 2025Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agentscited by 294
  5. 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 283
  6. 2025AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoderscited by 199
  7. 2025SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skillscited by 174
  8. 2025Dynamic Cheatsheet: Test-Time Learning with Adaptive Memorycited by 146
  9. 2024Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinementcited by 111
  10. 2024Automatically Interpreting Millions of Features in Large Language Modelscited by 106
  11. 2025SAEs Are Good for Steering -- If You Select the Right Featurescited by 74
  12. 2025Enhancing Automated Interpretability with Output-Centric Feature Descriptionscited by 39
  13. 2025Beyond the Surface: Measuring Self-Preference in LLM Judgmentscited by 30
  14. 2026Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Modelscited by 15
  15. 2025LinguaLens: Towards Interpreting Linguistic Mechanisms of Large Language Models via Sparse Auto-Encodercited by 13
  16. 2025SAGE: An Agentic Explainer Framework for Interpreting SAE Features in Language Modelscited by 10

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Do Sparse Autoencoders Identify Reasoning Features in Language Models?
  2. 2025Resurrecting the Salmon: Rethinking Mechanistic Interpretability with Domain-Specific Sparse Autoencoders
  3. 2025I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders
  4. 2025Probing the Representational Power of Sparse Autoencoders in Vision Models
  5. 2026Evaluating the Interpretability of Sparse Autoencoders with Concept Annotations
  6. 2025Rethinking Evaluation of Sparse Autoencoders through the Representation of Polysemous Words
  7. 2025Evaluating Sparse Autoencoders for Monosemantic Representation
  8. 2026SAEExplainer: Interpreting SAE Features with Activation-Guided Preference Optimization