AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Concept-TRAK: Understanding how diffusion models learn concepts through concept attribution

Yong-Hyun Park, Chieh-Hsin Lai, Satoshi Hayakawa, Yuhta Takida, Naoki Murata, Wei-Hsiang Liao, Woosung Choi, Kin Wai Cheuk, Junghyun Koo, Yuki Mitsufuji

ICLR 2026 · Image generation › Text-to-image diffusion · approach: Multi-concept composition

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 concept-level local attribution instead of image-level attribution on unseen concept combinationsConcept-TRAK · Synthetic dataset, out-of-distributionPrecision@10: baselines <= 0.50 -> 0.85Improved
use concept-level attribution on CelebA-HQ compositional OOD caseConcept-TRAK vs DAS · CelebA-HQ, out-of-distributionPrecision@10: 0.67 -> 0.97Improved
concept-level attribution in the in-distribution CelebA-HQ caseConcept-TRAK vs DAS · CelebA-HQ, in-distributionPrecision@10: 0.96 -> 0.92No clear effect
add concept-aware utility gradients, DPS-based training gradients, DDIM inversion, and gradient normalization progressively to TRAK with DSM loss (ablation)TRAK · AbC benchmark (48 samples)Recall@10: full method 0.955Improved
set-level (global) attribution by averaging utility gradients, compared with baselines on Toy color conceptConcept-TRAK · Set-level attribution, Toy dataset (average)Precision@10: D-TRAK 1.00 vs Concept-TRAK 0.95Hurt
average utility gradients over 256 generated images on CelebA-HQ set-level attributionbaselines · Set-level attribution, CelebA-HQPrecision@10: 1.00 for all methodsNo clear effect
concept-level local attribution for rare similar concepts (jaguar vs cat)Concept-TRAK · Similar big cats, local attributionPrecision@10: cat 0.925, jaguar 0.188Mixed
concept-level attribution on compositional promptsConcept-TRAK · Compositional common objects and styles, local objectPrecision@10: 0.025 -> 0.934Improved

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