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.

Mechanistic Insights into LoRA: Layer Sparsity for Adaptive Fine-Tuning via Path Patching
Oct 26, 2026Fei Zuo, Yizhou Huang, Kezhi WangLow-Rank Adaptation Parameter-Efficient Fine-Tuning
East China Normal University · Brunel University of London
Cross-Loop Query Handoff: How Later Loops Redirect Two-Hop Retrieval in the Ouro Looped Language Models
Oct 25, 2026Ziyu Mu, Zehua WangMulti-Hop Retrieval Multi-Hop QA
Faculty of Science The University of British Columbia · Department of Electrical and Computer Engineering The University of British Columbia
Beyond the Opening: Tracing the First Token's Persistent Influence on Reasoning Length
Oct 25, 2026Ruochen Liu, Hao Yan, Ziluowen Luo +2Mechanistic Interpretability
The Hong Kong University of Science and Technology (Guangzhou) · Central South University, China
Context or Knowledge: How LLMs Attribute Their Own Answers
Oct 23, 2026Kirill Afendulev, Alexey Dontsov, Alla Chepurova +2Claim Attribution in Language Models Data Provenance in Language Models
MI-MIDI 2: From Observation to Control with Sparse Autoencoders in Text-to-MIDI Generation
Oct 23, 2026Jakub Poćwiardowski, Mateusz ModrzejewskiMechanistic Interpretability Sparse Autoencoders
Institute of Computer Science, Warsaw University of Technology, Warsaw, Poland
Interpreting Model Reasoning through Behavioral Counterfactuals
Oct 22, 2026Shawn Im, James Oldfield, Adel Bibi +1Causal Interventions in Language Models Faithfulness of Language Model Reasoning
University of Wisconsin–Madison · University of Oxford
Rethinking SAE Evaluation: An Atomic Interpretable Unit Framework Reveals Hidden Polysemanticity and Redundancy
Oct 21, 2026Riteng Zhang, Xiaoqian WangSparse Autoencoders Explainability Evaluation
Purdue University
Understanding the Inner Workings of a State-Space Foundation Model for Limit Order Books
Oct 20, 2026Valentin Mohl, Alexandre Bismuth, Aramis Marti-Shahandeh +6Limit Order Book Modeling State Representation Learning
British Open-ended Learning and Discovery Lab (BOLD), Department of Computer Science, University of Oxford, Oxford, United Kingdom · Harvard University, Cambridge, MA, United States · British Open-ended Learning and Discovery Lab (BOLD), Department of Engineering Science, University of Oxford, Oxford, United Kingdom +3
Convergent Performance, Divergent Mechanisms in LoRA-Adapted Vision Foundation Models for Brain MRI
Oct 20, 2026Cecilia Diana-Albelda, Fiona Kekwick, Raghav Mehta +6Vision Foundation Model Adaptation Low-Rank Adaptation
Department of Electronic Technology and Communications, Universidad Autónoma de Madrid, Madrid, Spain · Department of Computing, Imperial College London, London, United Kingdom · Department of Brain Sciences, Imperial College London, London, United Kingdom +3
Uncovering Audio-Text Alignment in Large Audio-Language Models through Cross-Modal Concept Neurons
Oct 20, 2026Tzu-Quan Lin, Chih-Kai Yang, Yi-Cheng Lin +2Audio-Text Alignment Audio-Language Models
Centre for Speech Technology Research, University of Edinburgh, United Kingdom · Graduate Institute of Communication Engineering, National Taiwan University, Taiwan · NTU Artificial Intelligence Center of Research Excellence (NTU AI-CoRE), Taiwan
Shaped SAEs: Expanding the Toolkit for Mechanistic Interpretability
Oct 19, 2026Oveys Delafrooz Noroozi, Bhagyashree Puranik, Upamanyu MadhowLLM Interpretability Sparse Autoencoders
University of California, Santa Barbara
The Culture Within: Towards Mechanistic Understanding of Cultural Reasoning in LLMs
Oct 19, 2026Samee Arif, Furkan Danisman, Angana Borah +6Cultural Bias in Language Models Cultural Knowledge in Language Models
Inside the Scorer: Mechanistic Evidence on Rubric Control in LLM-Based Automated Essay Scoring from 25 Base Models
Oct 19, 2026Chungche Huang, Po-Ya Angela Wang, Shu-Kai HsiehAutomated Essay Scoring Rubric-Based Evaluation
Graduate Institute of Linguistics National Taiwan University · Co-intelligence Humanities AI (CHAI) Future Lab National Taiwan University
Text Dominance in Paralinguistic Conflict: Tracing Suppression Through Audio LLM Circuits
Oct 19, 2026Ankur Banerjee, Vishnu Guntakala, Danity Pike +3Audio-Language Models Audio-Language Model Evaluation
Independent Researcher · Brown University · University of Birmingham
Cognitive Maps in Large Language Models
Oct 19, 2026Albert Qin, Jonathan Cohen, Gautam ReddyCognitive Modeling Mechanistic Interpretability
Department of Physics, Princeton University, Princeton, New Jersey · Princeton Neuroscience Institute, Princeton University, Princeton, New Jersey
Transformers Learn Two Competing Sorting Algorithms, and Only One Length-Generalizes
Oct 18, 2026Khashayar Gatmiry, Nathan W. Henry, Chloe H. Su +5Length Generalization Transformer Training Dynamics
UC Berkeley & Apple · UC Berkeley · Kempner Institute, Harvard University +5
Safety Kernel: Mechanistic Alignment for Cross-Lingual Defense in LLMs
Oct 17, 2026Chengyu Xie, Fei Shen, Shuyi Miao +5LLM Safety Language Model Robustness
School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China · Nanjing University of Science and Technology, Nanjing, China · School of Artificial Intelligence, Beihang University, Beijing, China +3
Beyond SAEs: Consistent Concept Discovery with Sparse Coding
Oct 16, 2026Jinho Choi, Hyesu Lim, Jaegul Choo +1Sparse Autoencoders Sparse Coding
KAIST · Helmholtz Munich
Verbalising Model's Computation for Chain-of-Thought Faithfulness Detection
Oct 16, 2026Yujing Wang, Yulan He, Wenya Wang +1CoT Faithfulness Faithfulness of Language Model Reasoning
King’s College London · Nanyang Technological University
Compression Precedes, Content Decides: Emergent Misalignment in Reward Models
Oct 16, 2026Dhruv Puri, Nikhil PrakashReward Modeling Emergent Misalignment
Northeastern University
Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics
Oct 15, 2026Jian Gu, Hongyu Zhang, Chunyang Chen +1LLM Interpretability Mechanistic Interpretability
Monash University · Chongqing University · Technical University of Munich
Null Controls Are Necessary but Not Sufficient: Evaluating Acceptance Guards for Scientific Discovery and Interpretability
Oct 15, 2026Tianyu WuScientific Discovery Automated Evaluation
Project Aether
Human Interventions as Labels: Localizing, Steering, and (Sometimes) Editing the Computations Behind LLM Agent Failures
Oct 15, 2026Cesare Zavattari, Alessandro Tommasi, Giuseppe PrencipeLLM Agents Failure Attribution for LLM Agents
Dipartimento di Informatica, Università di Pisa, Pisa, Italy
Translation Circuit Discovery and Targeted Optimization for Low-Resource Multilingual Translation
Oct 15, 2026Atai Waris Khan, Abdullah Mazhar, Raghu Kumar +1Multilingual Language Models Low-Resource MT
Indraprastha Institute of Information Technology Delhi
Behavioral Compositionality of Sparse Autoencoder Representations
Oct 15, 2026Haotian Ma, Ziye Mei, Frederic SalaSparse Autoencoders Compositional Representation Learning
Department of Computer Sciences, University of Wisconsin–Madison
Do Emergent Misalignment Models Share a Circuit?
Oct 14, 2026Muhammad Haseeb, Muhammad TahirEmergent Misalignment in Language Models Circuits in Language Models
Lahore University of Management Sciences (LUMS)
Interpreting Language Model Parameters
Oct 14, 2026Lucius Bushnaq, Dan Braun, Oliver Clive-Griffin +5Mechanistic Interpretability Transformer Interpretability
Goodfire · MATS · Independent
Sparse Autoencoders for Mechanistic Interpretability in NLP: A Critical Survey of Recoverability, Interpretability, and Utility
Oct 14, 2026Salima Lamsiyah, Adelaide Danilov, Oleksandr Marchenko Breneur +1Sparse Autoencoders LLM Interpretability
Department of Computer Science (DCS), University of Luxembourg
Sparse Transformer: a Controlled Testbed for Mechanistic Interpretability
Oct 14, 2026Anton Korznikov, Alexey Dontsov, Elena TutubalinaMechanistic Interpretability Neuron-Level Interpretability
Multi-Stream Residual Computation in Diffusion Language Models
Oct 13, 2026Ekaterina Alimaskina, Suren Kazarian, Gleb Molodtsov +1Diffusion Language Models Masked Diffusion Models
BRAIn Lab, Yandex Research

Oct 26, 2026J/K move · Enter open · S save

Mechanistic Insights into LoRA: Layer Sparsity for Adaptive Fine-Tuning via Path Patching

Fei Zuo, Yizhou Huang, Kezhi Wang

East China Normal University · Brunel University of London

Low-Rank Adaptation Parameter-Efficient Fine-Tuning Mechanistic Interpretability

Open paper AskPDF arXiv

Low-Rank Adaptation (LoRA) has emerged as the dominant paradigm for parameter-efficient fine-tuning of large language models. Standard all-layer LoRA applies adaptation throughout the transformer, despite heterogeneous contributions from its components. In this work, we conduct a systematic mechanistic interpretability analysis to understand how LoRA affects model internals. We observe that LoRA's adaptation effect is highly non-uniform, with significant component-wise and layer-wise sparsity. Most critically, we find that important layers are highly task-dependent, with limited overlap across tasks. Motivated by these insights, we propose Adaptive-LoRA, a calibration-driven method that automatically identifies task-specific layers via frozen-base zero-ablation, enabling efficient layer selection without exhaustive search. Extensive experiments on Qwen2.5-1.5B across 11 diverse reasoning tasks demonstrate that Adaptive-LoRA outperforms full-layer LoRA while using only 30% of layers, achieving a

3.5\times

reduction in adapter parameters. This work bridges mechanistic interpretability and practical efficiency, providing both fundamental insights and a principled approach for adaptive fine-tuning.