This organization hosts Explainable AI (XAI) methods written in the Julia programming language, with a focus on post-hoc, local input-space explanations of black-box models. In simpler terms, methods that try to answer the question "Which part of the input is responsible for the model's output?".
The ecosystem is organized into several packages. As a user, you only need to install the packages that implement the methods you want to use.
As a developer, you can make use of the XAIBase.jl interface to quickly implement or prototype new methods without having to write boilerplate code.
Our recommended starting point into the Julia-XAI ecosystem is the Getting started guide in the Julia-XAI documentation.
If you want to implement an XAI method, take a look at the common interface defined in XAIBase.jl.
We welcome all contributions to the Julia-XAI ecosystem. Please contact us if you want your package to be included in this organization.
| Name | Latest release | Status | Summary |
|---|---|---|---|
| ExplainableAI.jl |  |
  |
Collection of Explainable AI methods in Julia |
| RelevancePropagation.jl |  |
  |
Layerwise Relevance Propagation (LRP) and CRP for use with Flux.jl |
| XAIBase.jl |  |
  |
Core package defining the Julia-XAI interface |
| VisionHeatmaps.jl |  |
  |
Heatmaps for vision models |
| TextHeatmaps.jl |  |
  |
Heatmaps for language models |
