CircuitKIT : Circuit Discovery, Evaluation, and Application Toolkit for Mechanistic Interpretability
2026-07-21 • Machine Learning
Machine LearningComputation and LanguageEmerging Technologies
AI summaryⓘ
The authors introduce CircuitKIT, a new library designed to make it easier to study and work with circuits in machine learning models. Currently, analyzing these circuits requires using many separate tools and writing special prompts, which can be complicated. CircuitKIT connects all parts of the analysis process in one place and includes tools for discovering, evaluating, and modifying circuits. This helps researchers compare methods more easily and apply circuit analysis to more tasks. The authors also provide examples and documentation online to support users.
circuit analysismachine learning modelspruningfine-tuningdiscovery algorithmscircuit diagnosticsinterventionmodel explanationdeclarative interfaces
Authors
Pratinav Seth, Hem Gosalia, Aditya Kasliwal, Vinay Kumar Sankarapu
Abstract
Circuit analysis can support not only model explanation but also downstream interventions such as pruning, editing, steering, and selective fine-tuning. However, conducting such analyses currently requires stitching together separate implementations for discovery, evaluation, and intervention, as well as hand-authoring the contrastive prompts required by many discovery methods. This fragmentation makes methods difficult to compare and limits their application beyond canonical tasks. We introduce CircuitKIT, a source-available library that connects the circuit-analysis workflow through a typed, serializable representation. CircuitKIT provides a suite of discovery algorithms, declarative interfaces for mapping structured data into discovery tasks, complementary circuit diagnostics, and downstream application modules. Together, these components provide common infrastructure for conducting and comparing circuit analyses. The library, examples, notebooks, and documentation are released at https://github.com/Lexsi-Labs/CircuitKIT .