XWhy: eXplain Why¶
Explain black-box behaviour with a SMILE.
XWhy: eXplain Why is a Python library for model-agnostic local explainability. It uses SMILE—Statistical Model-agnostic Interpretability with Local Explanations—to perturb an input, observe changes in model behaviour, fit a local surrogate model, and report feature-level influence together with explanation-quality evidence.
Current package maturity
XWhy v0.0.3 is currently classified as pre-alpha. Tabular, Image and LLM explainers are implemented. Other capabilities are clearly labelled as under construction or coming soon.
Start here¶
- Install XWhy
- Generate your first explanation
- Choose the correct explainer
- Browse all explainers and their status
- Read the generated API reference
Capability overview¶
| Capability | Public component | Documentation status | Implementation status |
|---|---|---|---|
| Image classification | ImageClassificationExplainer |
Available | Available |
| Image generation and editing | Pix2PixExplainer prototype |
Under construction | Interface only |
| LLM prompt-response | LLMExplainer |
Available | Available |
| Tabular | TabularExplainer |
Under construction | Interface only |
| Text | TextExplainer |
Under construction | Interface only |
| Point cloud | PointCloudExplainer |
Under construction | Interface only |
| Time series | Planned | Coming soon | Not yet implemented |
| Multimodal | Planned | Coming soon | Not yet implemented |
Pix2Pix is documented as one image-editing model family within the broader image-generation explainability section.
What XWhy explanations mean¶
XWhy produces local, perturbation-based approximations of model behaviour around a selected input. An explanation can identify associations between input components and changes in model output, but it does not expose a model's private internal reasoning or prove causality.
Read limitations and responsible use before using explanations in safety-critical, medical, legal, financial, or high-impact decisions.