AI Paper Atlas

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

Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept Discovery

Sukrut Rao, Sweta Mahajan, Moritz Böhle, Bernt Schiele

ECCV 2024 · Label-efficient and robust learning › Recognition and long-tail · approach: Concept bottleneck models · 41 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
use a single task-agnostic SAE concept layer instead of task-specific LLM-generated concept setsDN-CBM · Places365 (ResNet-50)accuracy (%): 52.7 -> 53.5Improved
task-agnostic concepts vs LaBo task-specific concepts on CIFAR100DN-CBM · CIFAR100 (ResNet-50)accuracy (%): 69.1 vs 67.5Hurt
select a separate SAE per downstream dataset instead of a shared one (ablation)DN-CBM · CIFAR100 (ResNet-50)accuracy (%): 67.5 -> 69.2Improved
intervene to keep only bird concepts in the concept bottleneck (ablation)DN-CBM · Waterbirds-100overall accuracy (%): 82.8 -> 89.4Improved
intervene to keep only bird concepts (ablation)DN-CBM · Waterbirds-100 Landbird on Wateraccuracy (%): 71.3 -> 86.6Improved
intervene to remove bird concepts (ablation)DN-CBM · Waterbirds-100overall accuracy (%): 82.8 -> 60.8Hurt
use SAE nodes instead of CLIP neurons for concepts, named with same vocabulary (ablation)DN-CBM · user study, semantic consistencyconsistency and name accuracy ratingsImproved
raise the L1 sparsity coefficient of the SAE (ablation)SAE · reconstruction errorreconstruction errorMixed
unsupervised SAE concept naming vs CLIP retrieval on SUNAttributesDN-CBM · SUNAttributesJaccard index: 22.0 vs 18.3Hurt

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