AI Paper Atlas

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

Auxiliary Losses for Learning Generalizable Concept-based Models

Ivaxi Sheth, Samira Ebrahimi Kahou

NeurIPS 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Concept bottleneck models · 4 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
add concept orthogonal loss (COL) (ablation)Coop-CBM · CUBtask accuracy: 83.6 -> 84.1Improved
add COL (ablation)Coop-CBM · TILtask accuracy: 53.4 -> 54.2Improved
add COL (ablation)Joint CBM · CUBconcept accuracy: 93.2 -> 96.4Improved
add COL (ablation)Independent CBM · CUBconcept accuracy: 96.6 -> 97.2Mixed
add COL (ablation)Coop-CBM · CUB out-of-domain spurious backgroundaccuracy: 35.4 -> 36.2Improved
train with hard (clipped) concept values (ablation)Coop-CBM · CUBtask accuracy: 83.6 -> 82.1Hurt
clip concept predictions to hard labels at evaluation (ablation)Coop-CBM · CUBtask accuracy: 83.6 -> 83.0Hurt
clip concept predictions to hard labels at evaluation (ablation)Coop-CBM + COL · CUBtask accuracy: 84.1 -> 84.0No clear effect
raise all loss weights alpha, beta, gamma from 0.01 to 1.0 (ablation)Coop-CBM + COL · CUBtask accuracy: 84.1 -> 83.0Hurt
vary COL weight lambda from 0.05 to 10.0 (ablation)Coop-CBM + COL · CUBtask accuracy: 84.1 -> 83.6No clear effect
vary COL weight lambda from 0.05 to 1.0 (ablation)Coop-CBM + COL · TILtask accuracy: 54.2 -> 53.6No clear effect
duplicate 10% to 100% of concepts (ablation)Coop-CBM + COL · CUBtask accuracy: 83.9 -> 83.4No clear effect

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