Saved in:
Bibliographic Details
Main Authors: Tian, Wenqing, Mao, Hanyi, Liu, Zhaocheng, Zhang, Lihua, Liu, Qiang, Wu, Jian, Wang, Liang
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.21937
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918403645636608
author Tian, Wenqing
Mao, Hanyi
Liu, Zhaocheng
Zhang, Lihua
Liu, Qiang
Wu, Jian
Wang, Liang
author_facet Tian, Wenqing
Mao, Hanyi
Liu, Zhaocheng
Zhang, Lihua
Liu, Qiang
Wu, Jian
Wang, Liang
contents Subject-driven image generation is increasingly expected to support fine-grained control over multiple entities within a single image. In multi-reference workflows, users may provide several subject images, a background reference, and long, entity-indexed prompts to control multiple people within one scene. In this setting, a key failure mode is cross-subject attribute misbinding: attributes are preserved, edited, or transferred to the wrong subject. Existing benchmarks and metrics largely emphasize holistic fidelity or per-subject self-similarity, making such failures hard to diagnose. We introduce MultiBind, a benchmark built from real multi-person photographs. Each instance provides slot-ordered subject crops with masks and bounding boxes, canonicalized subject references, an inpainted background reference, and a dense entity-indexed prompt derived from structured annotations. We also propose a dimension-wise confusion evaluation protocol that matches generated subjects to ground-truth slots and measures slot-to-slot similarity using specialists for face identity, appearance, pose, and expression. By subtracting the corresponding ground-truth similarity matrices, our method separates self-degradation from true cross-subject interference and exposes interpretable failure patterns such as drift, swap, dominance, and blending. Experiments on modern multi-reference generators show that MultiBind reveals binding failures that conventional reconstruction metrics miss.
format Preprint
id arxiv_https___arxiv_org_abs_2603_21937
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MultiBind: A Benchmark for Attribute Misbinding in Multi-Subject Generation
Tian, Wenqing
Mao, Hanyi
Liu, Zhaocheng
Zhang, Lihua
Liu, Qiang
Wu, Jian
Wang, Liang
Computer Vision and Pattern Recognition
Subject-driven image generation is increasingly expected to support fine-grained control over multiple entities within a single image. In multi-reference workflows, users may provide several subject images, a background reference, and long, entity-indexed prompts to control multiple people within one scene. In this setting, a key failure mode is cross-subject attribute misbinding: attributes are preserved, edited, or transferred to the wrong subject. Existing benchmarks and metrics largely emphasize holistic fidelity or per-subject self-similarity, making such failures hard to diagnose. We introduce MultiBind, a benchmark built from real multi-person photographs. Each instance provides slot-ordered subject crops with masks and bounding boxes, canonicalized subject references, an inpainted background reference, and a dense entity-indexed prompt derived from structured annotations. We also propose a dimension-wise confusion evaluation protocol that matches generated subjects to ground-truth slots and measures slot-to-slot similarity using specialists for face identity, appearance, pose, and expression. By subtracting the corresponding ground-truth similarity matrices, our method separates self-degradation from true cross-subject interference and exposes interpretable failure patterns such as drift, swap, dominance, and blending. Experiments on modern multi-reference generators show that MultiBind reveals binding failures that conventional reconstruction metrics miss.
title MultiBind: A Benchmark for Attribute Misbinding in Multi-Subject Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.21937