Mechanistic Interpretability

Follow this topic Ask about this topic

The field seeks to explain how a model’s internal computations produce particular outputs, rather than relying only on input-output behavior. Researchers use sparse autoencoders and transcoders to identify features, and probes, attribution graphs, activation patching, ablation, and steering to trace or test their roles. Studies examine language-model tasks such as factual recall, reasoning, and refusal, as well as vision, audio, and scientific models. A central challenge is that a signal can be easy to decode without being causally used, while circuit explanations and interventions can vary across prompts, models, and analysis choices.

Latest papers

388 papers63 papers in the last four weeks, up 75% on the four weeks before. 0.4% of all new papers.

Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language Models
Mar 6, 2026Jinman Wu, Yi Xie, Shen Lin +2Refusal Mechanisms in Language Models LLM Safety Alignment
Xidian University, Xi’an, China · Tsinghua University, Shenzhen, China · Fujian Normal University, Fuzhou, China +1
Causal Circuit Tracing Reveals Distinct Computational Architectures in Single-Cell Foundation Models: Inhibitory Dominance, Biological Coherence, and Cross-Model Convergence
Mar 2, 2026Ihor KendiukhovMechanistic Interpretability Causal Interpretability
Department of Computer Science, University of Tübingen, Tübingen, Germany
CIRCUS: Circuit Consensus under Uncertainty via Stability Ensembles
Feb 28, 2026Swapnil ParekhCircuit Discovery Circuits in Language Models
Intuit
GLUScope: A Tool for Analyzing GLU Neurons in Transformer Language Models
Feb 27, 2026Sebastian Gerstner, Leonor Veloso, Hinrich SchützeNeuron-Level Interpretability Transformer Interpretability
LMU Munich and Munich Center for Machine Learning (MCML), Munich, Germany
A Case Study on Concept Induction for Neuron-Level Interpretability in CNN
Feb 27, 2026Moumita Sen Sarma, Samatha Ereshi Akkamahadevi, Pascal HitzlerNeuron-Level Interpretability Neural Network Interpretability
Department of Computer Science, Kansas State University, Manhattan, Kansas, USA
Certified Circuits: Stability Guarantees for Mechanistic Circuits
Feb 26, 2026Alaa Anani, Tobias Lorenz, Bernt Schiele +2Circuit Discovery Mechanistic Interpretability
Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbrücken, Germany · CISPA Helmholtz Center for Information Security, Saarbrücken, Germany
MINAR: Mechanistic Interpretability for Neural Algorithmic Reasoning
Feb 24, 2026Jesse He, Helen Jenne, Max Vargas +4Neural Algorithmic Reasoning Mechanistic Interpretability
Pacific Northwest National Laboratory, Richland, WA · Halıcı˘oğlu Data Science Institute, University of California, San Diego, San Diego, CA · Department of Computer and Information Science, University of Pennsylvania, Pennsylvaina, PA +1
AR&D: A Framework for Retrieving and Describing Concepts for Interpreting AudioLLMs
Feb 24, 2026Townim Faisal Chowdhury, Ta Duc Huy, Siqi Pan +2Multimodal Model Interpretability Mechanistic Interpretability
Australian Institute for Machine Learning, University of Adelaide · Dolby Laboratories · School of Computer and Mathematical Sciences, University of Adelaide, Australia
Simplifying Outcomes of Language Model Component Analyses with ELIA
Feb 20, 2026Aaron Louis Eidt, Nils FeldhusNatural Language Explanation Generation LLM Interpretability
Technische Universität Berlin · Fraunhofer Heinrich Hertz Institute · BIFOLD – Berlin Institute for the Foundations of Learning and Data
Systematic Evaluation of Single-Cell Foundation Model Interpretability Reveals Attention Captures Co-Expression Rather Than Unique Regulatory Signal
Feb 19, 2026Ihor KendiukhovMechanistic Interpretability Neural Network Interpretability
Department of Computer Science, University of Tübingen, Tübingen, Germany
Discovering Universal Activation Directions for PII Leakage in Language Models
Feb 19, 2026Leo Marchyok, Zachary Coalson, Sungho Keum +2Privacy Leakage in Language Models LLM Interpretability
Oregon State University, Corvallis OR, USA · Korea Advanced Institute of Science & Technology, Daejeon, South Korea.
Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable Guarantees
Feb 18, 2026Itamar Hadad, Guy Katz, Shahaf BassanCircuit Discovery Mechanistic Interpretability
School of Computer Science and Engineering, The Hebrew University of Jerusalem
Delta-Crosscoder: Robust Crosscoder Model Diffing in Narrow Fine-Tuning Regimes
Feb 16, 2026Aly Kassem, Thomas Jiralerspong, Negar Rostamzadeh +1Mechanistic Interpretability
Mila, Quebec AI Institute, Quebec, Canada · Université de Montréal, Quebec, Canada · McGill University, Quebec, Canada
Two Stages of Folding: Convergent Mechanisms in AI Protein Folding Trunks
Feb 5, 2026Kevin Lu, Jannik Brinkmann, Stefan Huber +4Protein Structure Prediction AI-Assisted Scientific Research
1Northeastern University · 2TU Clausthal · 3Harvard University +2
Towards Isolated Interventions via Almost Orthogonal Features in Language Models
Feb 4, 2026Moritz Miller, Florent Draye, Bernhard SchölkopfDisentangled Representation Learning Representation Learning
Max Planck Institute for Intelligent Systems, Tübingen, Germany · ETH Zurich, Switzerland · Max Planck ETH Center for Learning Systems +1
Universal Redundancies in Time Series Foundation Models
Feb 2, 2026Anthony Bao, Venkata Hasith Vattikuti, Jeffrey Lai +1Transformer Interpretability Kernel Regression
ECE Department, UT Austin · Department of Physics, UT Austin · Oden Institute, UT Austin
Mechanistic Evidence for Spectral Structures in Prior-Data Fitted Networks
Jan 29, 2026Kaustubh Sharma, Srijan Tiwari, Ojasva Nema +1Prior-Data Fitted Networks Neural Network Interpretability
Indian Institute of Technology Roorkee, India
Mechanism Shift During Post-training from Autoregressive to Masked Diffusion Language Models
Jan 21, 2026Injin Kong, Hyoungjoon Lee, Yohan JoMasked Diffusion Models Autoregressive Diffusion
Graduate School of Data Science, Seoul National University · Department of Biosystems & Biomaterials Science and Engineering, Seoul National University
Tracing the Latent Threads: A Mechanistic Study of How LLMs Represent and Operationalize Race and Ethnicity Cues
Jan 19, 2026Shiyue Hu, Ruizhe Li, Yanjun GaoLLM Interpretability Mechanistic Interpretability
University of Colorado Anschutz · University of Colorado Boulder · University of Aberdeen
Beyond Transfer Accuracy: Mechanism-Guided Controlled Adaptation for Low-Resource Languages
Jan 13, 2026Khumaisa Nur'aini, Ayu Purwarianti, Alham Fikri Aji +1Selective Fine-Tuning Circuit Discovery
Monash University Indonesia · MBZUAI · Institute Teknologi Bandung +1
Emergence of Minimal Circuits for Indirect Object Identification in Attention-Only Transformers
Oct 28, 2025Rabin AdhikariAttention Head Analysis Transformer
Saarland University 66123 Saarbrücken
Attribution Graphs and Causal Probing for Mechanistic Discovery and Bias Repair in Multimodal Generative Learning
Oct 14, 2025Noor Islam S. Mohammad, Uluğ BayazıtGradient-Based Attribution Causal Representation Learning
1,2Istanbul Technical University
Circuit Distillation
Sep 29, 2025Somin Wadhwa, Silvio Amir, Byron C. WallaceLanguage Model Distillation Mechanistic Interpretability
Khoury College of Computer Sciences, Northeastern University, Boston, MA 02115, USA
Interpreting vision transformers via residual replacement model
Sep 22, 2025Jinyeong Kim, Junhyeok Kim, Yumin Shim +3Transformer Interpretability Vision Transformer
Yonsei University
Towards Inclusive Toxic Content Moderation: Addressing Vulnerabilities to Adversarial Attacks in Toxicity Classifiers Tackling LLM-generated Content
Sep 16, 2025Shaz Furniturewala, Arkaitz ZubiagaMechanistic Interpretability Toxicity Detection
Center for Data Science, New York University · Queen Mary University of London
Interpretability and Generalization Bounds for Learning Spatial Physics
Jun 18, 2025Alejandro Francisco Queiruga, Theo Gutman-Solo, Shuai JiangNeural Network Generalization Mechanistic Interpretability
OpenAI, San Francisco, CA, USA · Google, Mountain View, CA, USA · Sandia National Laboratories, Albuquerque, NM, USA
Kronecker Factorization Improves Efficiency and Interpretability of Sparse Autoencoders
May 28, 2025Vadim Kurochkin, Yaroslav Aksenov, Daniil Laptev +2Representation Learning Activation Sparsity
T-Tech
Interpreting Global Perturbation Robustness of Image Models using Axiomatic Spectral Importance Decomposition
Aug 2, 2024Róisín Luo, James McDermott, Colm O'RiordanImage Corruption Robustness Neural Network Robustness
SFI Centre for Research Training in Artificial Intelligence · School of Computer Science, University of Galway

Mar 6, 2026J/K move · Enter open · S save

Knowing without Acting: The Disentangled Geometry of Safety Mechanisms in Large Language Models

Jinman Wu, Yi Xie, Shen Lin, Shiqian Zhao+1

Xidian University, Xi’an, China · Tsinghua University, Shenzhen, China · Fujian Normal University, Fuzhou, China+1

Refusal Mechanisms in Language Models LLM Safety Alignment LLM Refusal LLM Jailbreak Attacks Mechanistic Interpretability LLM Safety Benchmarks

Open paper AskPDF arXiv

Safety alignment is often conceptualized as a monolithic process wherein harmfulness detection automatically triggers refusal. However, the persistence of jailbreak attacks suggests a fundamental mechanistic decoupling. We propose the Disentangled Safety Hypothesis (DSH), positing that safety computation operates on two distinct subspaces: a Recognition Axis (

\mathbf{v}_H

, Knowing'') and an *Execution Axis* ($\mathbf{v}_R$, Acting''). Our geometric analysis reveals a universal

Reflex-to-Dissociation'' evolution, where these signals transition from antagonistic entanglement in early layers to structural independence in deep layers. To validate this, we introduce *Double-Difference Extraction* and *Adaptive Causal Steering*. Using our curated AmbiguityBench, we demonstrate a causal double dissociation, effectively creating a state of

Knowing without Acting.'' Crucially, we leverage this disentanglement to propose the Refusal Erasure Attack (REA), which achieves State-of-the-Art attack success rates by surgically lobotomizing the refusal mechanism. Furthermore, we uncover a critical architectural divergence, contrasting the Explicit Semantic Control of Llama3.1 with the Latent Distributed Control of Qwen2.5. The code and dataset are available at https://anonymous.4open.science/r/DSH.