Quick start¶
XWhy currently provides five executable explainer areas. Choose the workflow that matches the behaviour you want to explain.
Explain an image classifier¶
Use ImageClassificationExplainer for a PyTorch image-classification model. The explainer estimates how image regions influence a selected prediction.
Open the image-classification tutorial
Explain image generation or editing¶
Use ImageGenerationAndEditingExplainer for supported image-generation or image-editing providers, pipelines, or compatible custom generation functions. The workflow perturbs the textual instruction, generates or edits images, measures changes in the image output, and fits a local surrogate model.
Open the image generation & editing guide
Explain an LLM response¶
Use LLMExplainer to perturb a text prompt, compare the resulting model responses, and estimate local word influence.
Explain a tabular prediction¶
Use TabularExplainer for structured classification or regression. The implementation generates local perturbations, computes feature-distribution distances, queries the black-box model, and fits a weighted surrogate explanation.
Open the tabular explainer guide
Explain a conventional text prediction¶
Use TextExplainer for text classifiers or compatible prediction functions. It perturbs the input text, evaluates the black-box model, computes text distance, and estimates local word contributions.
Development and roadmap capabilities¶
PointCloudExplainer is currently an exported development interface whose explain() method raises NotImplementedError.
Time Series, Multimodal, Agentic AI, and Multi-Agent AI are documented roadmap capabilities and are not yet exported as supported explainers.
See the explainer status matrix before designing a workflow.