AI Paper Atlas

Research map / Large language models / AI and society

Attribution Analysis-based Concept Alignment: A Human-in-the-loop Data Debugging Framework

Lei Chai, Lu Qi, Hailong Sun, Jing Zhang, Jingxuan Xu

AAAI 2026 · Large language models › AI and society · approach: Benchmarking and reproducibility practice · 0 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 priority-score selection (entropy and low confound similarity) of synthetic images (ablation)AACA app. · Places365 subset, 1500 debug imagesaverage accuracy (%): 86.44 -> 89.92Improved
add priority-score selection of synthetic images (ablation)AACA app. · Places365 subset, 2000 debug imagesaverage accuracy (%): 85.25 -> 89.04Improved
add AACA sel. debug images to training setMAE baseline · Places365 subsetaverage accuracy (%): 86.75 -> 89.92Improved
add AACA app. (unselected) debug images to training setMAE baseline · Places365 subsetaverage accuracy (%): 86.75 -> 85.25Mixed
use human-annotated concepts instead of GPT-4 concepts for prompts (ablation)AACA sel. · Places365 subset, 1500 debug imagesaverage accuracy (%): 87.42 -> 89.92Improved
use AACA sel. instead of active-learning selection of search-based imagesMAE baseline · Places365 subset, 1500 debug imagesaverage accuracy (%): 88.08 -> 89.92Improved
swap text-to-image model to FLUX.1 (ablation)AACA sel. · Places365 subset, 1500 debug imagesaverage accuracy (%): 88.52 -> 89.92Improved
use AACA sel. with different T2I models (ablation)AACA sel. · Places365 subsetaverage accuracy (%): 86.75 -> 89.08Improved
add AACA debug images to training setDinit · Places365 subset concept distributionJensen-Shannon divergence: 0.152836 -> 0.084219Improved
add AACA debug images to training setDinit · Human evaluation of concept relevanceLikert score: 3.060 -> 3.520Improved
add randomly selected search imagesMAE baseline · Places365 subsetaverage accuracy (%): 86.75 -> 86.92No 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