AI Paper Atlas

Research map / Learning theory and algorithms / Fairness

SMiLE: Provably Enforcing Global Relational Properties in Neural Networks

Matteo Francobaldi, Michele Lombardi, Andrea Lodi

AAAI 2026 · Learning theory and algorithms › Fairness · approach: Post-processing and transport for fairness · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
enforce global robustness with SMiLE instead of property-agnostic trainingbase CNN · MNIST zero-vs-rest, delta=0.100PGD accuracy (%): 66.80 -> 92.17Improved
use SMiLE instead of CROWN-IBP for robustnessCROWN-IBP · MNIST zero-vs-rest, delta=0.100verified accuracy (%): 99.00 -> 90.57Hurt
use SMiLE instead of CROWN-IBPCROWN-IBP · MNISTtraining time (s): 14,138s -> 7,717sImproved
use SMiLE instead of CROWN-IBPCROWN-IBP · MNIST test-set predict and verifyinference time (s): 11.20s -> 0.03sImproved
enforce fairness with SMiLE instead of CertiFairCertiFair · Compas, Law, Crimeaccuracy at equal or better fairnessImproved
enforce fairness with SMiLE instead of property-agnostic trainingAgnostic · Compas, Law, Crimecounterfactual variationImproved
enforce monotonicity with SMiLE on non-monotonic synthetic functionsSMiLE · 9 synthetic monotonic tasksR2: 0.99 down to 0.83Mixed
replace original SMiLE pretraining with box-degeneracy-aware pretraining lossSMiLE · relational propertiestraining stabilityImproved
replace PGD training loop of original SMiLE with dual ascent and unrestricted projectionSMiLE · relational propertiesconvergence and generator callsImproved
use timeout-extension scheme in counterexample generation instead of complete searchSMiLE generator · relational propertiestraining timeImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy