MURANO: Design, Run, and Reproduce Mechanistic Interpretability Experiments as Composable Pipelines
cs.CL
Submitted: 2026-08-31
Updated: 2026-08-31
Comments: Accepted to the EMNLP 2026 System Demonstrations Track. 11 pages, 6 figures, 2 tables
Code: https://github.com/UKPLab/murano
Project page: https://ukplab.github.io/murano/https://pypi.org/project/murano-interp
License: http://creativecommons.org/licenses/by/4.0/
The gist: This paper presents Murano, an open source framework for designing, running, and reproducing mechanistic interpretability studies of large language models, intended for researchers across disciplines.
Terminology
Abstract
This paper presents Murano, an open source framework for designing, running, and reproducing mechanistic interpretability studies of large language models, intended for researchers across disciplines. These studies often combine loading, recording, attribution, intervention, and evaluation, while existing libraries tend to focus on different parts of this workflow. As a result, researchers using several libraries may need to adapt outputs from one for use by another. To bridge this gap, Murano represents operations from these five areas as composable steps. Steps exchange named result artifacts and declare the inputs they require and the outputs they produce. A pipeline executes its steps in the order supplied, and Murano uses canonical addresses when component identities pass between operations. Murano builds on existing interpretability and machine learning libraries. We demonstrate Murano through two reproductions of established interpretability studies and one illustrative sparse autoencoder case study.
Sources
- nnterp: A Standardized Interface for Mechanistic Interpretability of Transformers
- NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals
- Localizing Model Behavior with Path Patching
- A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models
- Gemma 2: Improving Open Language Models at a Practical Size
- Llama 2: Open Foundation and Fine-Tuned Chat Models
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering