AI Paper Atlas

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

Overlooked Factors in Concept-Based Explanations: Dataset Choice, Concept Learnability, and Human Capability

Vikram V. Ramaswamy, Sunnie S. Y. Kim, Ruth Fong, Olga Russakovsky

CVPR 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Post-hoc explanation analysis · 35 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
Switch the probe dataset between ADE20k and PascalBaseline, NetDissect, and TCAV explanations for a Places365 scene classifier · ADE20k vs. Pascal probe datasetsexplanation similarity: 0.078 average cosine similarityHurt
Use the Pascal probe dataset rather than ADE20kNetDissect neuron labels · 123 labeled neuronslabel agreement: 69/123Hurt
Use Broden concepts as prediction targets instead of Places365 classesLinear concept predictors · Broden concepts vs. Places365 classesmedian normalized AP: 37.5% -> 7.6%Hurt
Use CUB concepts as prediction targets instead of CUB classesLinear concept predictors · CUB concepts vs. CUB classesmedian normalized AP: 65.9% -> 2.3%Hurt
Show 32 rather than 8 conceptsBaseline explanations in the human study · Human concept recognitionmean recall: 71.7% ± 27.7% -> 56.8% ± 24.9%Hurt
Show more concepts, from 8 to 16 or 32Baseline explanations in the human study · Human model-output predictionmean accuracy: 64.8% ± 23.9% (8 concepts), 63.2% ± 26.9% (16 concepts), 63.6% ± 22.2% (32 concepts)Mixed
Use example-based explanations instead of 8-concept explanationsHuman model-output prediction · Human model-output predictionmean accuracy: 64.8% ± 23.9% vs. 60.0% ± 30.2%No clear effect
Limit explanations to 8, 16, or 32 concepts rather than 4 or 64Concept explanations in the preference study · Human preference for explanation complexityparticipants preferring the selected concept counts: 82%Improved

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