Right for the Right Reasons: masked input-gradient regularization

What changed

Adds annotation-weighted input-gradient penalties to prediction loss so models avoid selected irrelevant features, and uses gradient constraints to search for alternative decision boundaries.

Scientific scope

Adds annotation-weighted input-gradient penalties to prediction loss so models avoid selected irrelevant features, and uses gradient constraints to search for alternative decision boundaries. Local derivative penalties depend on annotation quality and differentiability; they do not guarantee global causal correctness or arbitrary distribution-shift robustness. General ML method only, not the clinical motivation or applications.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Local derivative penalties depend on annotation quality and differentiability; they do not guarantee global causal correctness or arbitrary distribution-shift robustness. General ML method only, not the clinical motivation or applications.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Right for the Right Reasons: Training Differentiable Models by Constraining their Explanations

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction