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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Park, Yonghyun, Lai, Chieh-Hsin, Hayakawa, Satoshi, Takida, Yuhta, Murata, Naoki, Liao, Wei-Hsiang, Choi, Woosung, Cheuk, Kin Wai, Koo, Junghyun, Mitsufuji, Yuki
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918364826304512
author Park, Yonghyun
Lai, Chieh-Hsin
Hayakawa, Satoshi
Takida, Yuhta
Murata, Naoki
Liao, Wei-Hsiang
Choi, Woosung
Cheuk, Kin Wai
Koo, Junghyun
Mitsufuji, Yuki
author_facet Park, Yonghyun
Lai, Chieh-Hsin
Hayakawa, Satoshi
Takida, Yuhta
Murata, Naoki
Liao, Wei-Hsiang
Choi, Woosung
Cheuk, Kin Wai
Koo, Junghyun
Mitsufuji, Yuki
contents While diffusion models excel at image generation, their growing adoption raises critical concerns about copyright issues and model transparency. Existing attribution methods identify training examples influencing an entire image, but fall short in isolating contributions to specific elements, such as styles or objects, that are of primary concern to stakeholders. To address this gap, we introduce concept-level attribution through a novel method called Concept-TRAK, which extends influence functions with a key innovation: specialized training and utility loss functions designed to isolate concept-specific influences rather than overall reconstruction quality. We evaluate Concept-TRAK on novel concept attribution benchmarks using Synthetic and CelebA-HQ datasets, as well as the established AbC benchmark, showing substantial improvements over prior methods in concept-level attribution scenarios. We further demonstrate its versatility on real-world text-to-image generation with compositional and multi-concept prompts.
format Preprint
id arxiv_https___arxiv_org_abs_2507_06547
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Concept-TRAK: Understanding how diffusion models learn concepts through concept-level attribution
Park, Yonghyun
Lai, Chieh-Hsin
Hayakawa, Satoshi
Takida, Yuhta
Murata, Naoki
Liao, Wei-Hsiang
Choi, Woosung
Cheuk, Kin Wai
Koo, Junghyun
Mitsufuji, Yuki
Computer Vision and Pattern Recognition
Machine Learning
While diffusion models excel at image generation, their growing adoption raises critical concerns about copyright issues and model transparency. Existing attribution methods identify training examples influencing an entire image, but fall short in isolating contributions to specific elements, such as styles or objects, that are of primary concern to stakeholders. To address this gap, we introduce concept-level attribution through a novel method called Concept-TRAK, which extends influence functions with a key innovation: specialized training and utility loss functions designed to isolate concept-specific influences rather than overall reconstruction quality. We evaluate Concept-TRAK on novel concept attribution benchmarks using Synthetic and CelebA-HQ datasets, as well as the established AbC benchmark, showing substantial improvements over prior methods in concept-level attribution scenarios. We further demonstrate its versatility on real-world text-to-image generation with compositional and multi-concept prompts.
title Concept-TRAK: Understanding how diffusion models learn concepts through concept-level attribution
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2507.06547