CircuitKIT : Circuit Discovery, Evaluation, and Application Toolkit for Mechanistic Interpretability
Merged summary
TL;DR - CircuitKIT is a source-available toolkit that unifies circuit discovery, evaluation, and model intervention. Its shared infrastructure aims to make circuit-analysis methods easier to compare and apply beyond standard tasks.
- Uses a typed, serializable representation across the circuit-analysis workflow.
- Includes multiple discovery algorithms and declarative tools for generating discovery tasks from structured data.
- Provides circuit diagnostics and modules for pruning, editing, steering, and selective fine-tuning.
- Releases the library alongside examples, notebooks, and documentation.
Sources (1)
CircuitKIT : Circuit Discovery, Evaluation, and Application Toolkit for Mechanistic Interpretability
TL;DR - CircuitKIT is a source-available toolkit that unifies circuit discovery, evaluation, and model intervention. Its shared infrastructure aims to make circuit-analysis methods easier to compare and apply beyond standard tasks.
- Uses a typed, serializable representation across the circuit-analysis workflow.
- Includes multiple discovery algorithms and declarative tools for generating discovery tasks from structured data.
- Provides circuit diagnostics and modules for pruning, editing, steering, and selective fine-tuning.
- Releases the library alongside examples, notebooks, and documentation.