AI Paper Atlas

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

Leakage and Interpretability in Concept-Based Models

Enrico Parisini, Tapabrata Chakraborti, Chris Harbron, Ben D. MacArthur, Christopher R.S. Banerji

JMLR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Concept bottleneck models

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
use CTL instead of OIS as leakage indicator (ablation)leakage scores · TabularToy(0.25)Pearson r with S_int: 0.75 vs 0.49Improved
use CTL instead of OIS as leakage indicator (ablation)leakage scores · HAM10KPearson r with S_int: 0.49 vs -0.06Improved
use CTL instead of OIS as leakage indicator (ablation)leakage scores · dSprites(4)Pearson r with S_int: 0.67 vs 0.55Improved
use ICL instead of OIS (ablation)leakage scores · 3dshapes(0)Pearson r with S_int: 0.86 vs 0.82Improved
increase concept supervision lambda in soft/logit CBMs (ablation)soft and logit CBMs · TabularToy, dSprites, 3dshapes, CUB, HAM10KCTL and ICL leakage: leakage decreases as concept supervision increasesImproved
increase lambda beyond about 10 (ablation)soft and logit CBMs · TabularToy(0.25)task accuracy: task accuracy above 90% in only 3 out of 5 folds for lambda = 10, 20, 50Hurt
use logit instead of soft probability concept encoding for binary concepts (ablation)CBMs · several datasetsminimal attainable leakage: minimal leakage higher for logit CBMs than soft CBMsHurt
complete the annotated concept set (vs incomplete) (ablation)soft CBMs · 15 model classesCTL/ICL leakage: more leakage in 14/15 model classes with incomplete conceptsImproved
use a flexible final head instead of a misspecified linear head (ablation)soft CBMs · TabularToy, dSprites, 3dshapesCTL/ICL leakage: misspecified head causes higher leakage in 7/9 model classesImproved
incomplete concept set lowers reference task accuracy (ablation)final head · 3dshapes(0)y_acc: 1.000 -> 0.655Hurt
non-linear task with linear head (ablation)soft CBM linear head · TabularToy(0.25)y_acc: 1.000 -> 0.687Hurt
increase lambda and p_int in CEMs (ablation)CEMs · TabularToy, CUB and othersinterconcept leakage (vectors): interconcept information in vectors grows with lambda and p_intHurt

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