AI Paper Atlas

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

Probabilistic Concept Bottleneck Models

Eunji Kim, Dahuin Jung, Sangha Park, Siwon Kim, Sungroh Yoon

ICML 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Concept bottleneck models · 6 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
introduce probabilistic rather than deterministic embeddings (ablation)ProbCBM · CUB, 224 × 224Class accuracy: 0.677$\pm$0.004 -> 0.680$\pm$0.004No clear effect
use embedding means instead of Monte Carlo sampling at inference (ablation)ProbCBM · CUB, 224 × 224Class accuracy: 0.679$\pm$0.003 -> 0.680$\pm$0.004No clear effect
use probabilistic embeddings instead of deterministic embeddings (ablation)ProbCBM · AwA2Class accuracy: 0.880$\pm$0.002 -> 0.880$\pm$0.002No clear effect
use embedding means instead of Monte Carlo sampling at inference (ablation)ProbCBM · AwA2Class accuracy: 0.880$\pm$0.001 -> 0.880$\pm$0.002No clear effect
increase class-training concept replacement probability to 1 (ablation)ProbCBM · CUBClassification performanceHurt
vary the concept embedding dimension (ablation)ProbCBM · CUBConcept and class prediction performanceNo clear effect
vary the class embedding dimension (ablation)ProbCBM · CUBConcept and class prediction performanceNo clear effect
use a small concept embedding dimension (Dc=4) (ablation)ProbCBM · CUB occlusion uncertainty detectionProportion of samples with increased uncertainty: Dc=4Hurt
add ground-truth concept replacement during class-predictor trainingProbCBM · CUB and AwA2Class accuracyImproved
rank concepts by uncertainty for intervention rather than intervening in random orderProbCBM · CUBClass accuracyImproved
occlude image regionsProbCBM · CUB and AwA2Uncertainty detectionImproved

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