Accurate locally additive models with conditional feature effects for tabular data. Python package for the NeurIPS 2026 paper Interpretability-by-Design with Accurate Locally Additive Models and Conditional Feature Effects (arXiv).
Project page: https://givasile.github.io/calm-additive
The package is being released in October 2026. Until then this repository holds the project page only.
from calm_additive import CALMRegressor
model = CALMRegressor().fit(X_train, y_train)
model.predict(X_test)
model.explain()Built on effector.