AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Hide&Seek: Learning to Explain in an End-to-End Differentiable Network

Tal Ellinson, Hadi Mohasel Afshar, Sally Cripps

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 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
Replace joint INVASE selector-predictor with Hide&SeekSelector-predictor IWFS · Synthetic Syn4TPR (%): 90 -> 99Improved
Use Hide&Seek instead of REAL-xSelector-predictor IWFS · Synthetic Syn5TPR (%): 88 -> 97Improved
Use Hide&Seek instead of INVASE for training timeSelector-predictor IWFS · Synthetic data run timerun time (hh:mm:ss): 01:18:52 -> 00:00:05Improved
Replace zero/mean ablation with random marginal draws and continuous masks (vs INVASE) (ablation)Hide&Seek · Switch accuracy Syn4-6switch accuracy: 51 -> 100Improved
Replace zero/mean ablation with random marginal draws (vs L2X) (ablation)Hide&Seek · Switch accuracy Syn4-6switch accuracy: 58 -> 100Improved
Correct INVASE's failures to identify the switch feature (ablation)INVASE · Synthetic Syn4TPR (%): 90 -> 99.9Improved
Add high feature correlation rho up to 0.9Hide&Seek · Correlated Syn4-6switch accuracy: 0.998 -> 0.994No clear effect
Use Hide&Seek instead of INVASE under high feature correlationSelector-predictor IWFS · Credit default semi-synthetic Syn4F1 (%): 57 -> 74Improved
Use Hide&Seek instead of INVASE under high feature correlationSelector-predictor IWFS · Credit default semi-synthetic Syn2F1 (%): 100 -> 67Hurt
Use Hide&Seek instead of INVASE on real credit default labelsSelector-predictor · Credit default CDY (real labels)AUROC: 0.771 -> 0.770No clear effect
Add parsimony-weight annealing (quadratic growth of lambda) (ablation)Hide&Seek · Synthetic datafeature importance metricsImproved
Use Hide&Seek instead of INVASE for AUROC on Syn1 credit labelsSelector-predictor · Credit default semi-synthetic Syn1 labelsAUROC: 0.601 -> 0.664Improved

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