I2I-Bench：面向图像到图像编辑模型的综合基准套件
I2I-Bench: A Comprehensive Benchmark Suite for Image-to-Image Editing Models
做图像编辑模型时可直接用这套含 10 类任务、30 个解耦维度的自动化评测协议（专用工具 + LMM 混合打分）替代人工评测，并用其对齐验证方法检验自建指标是否与人类偏好一致。
原文依据 30 decoupled and fine-grained evaluation dimensions with automated hybrid evaluation methods that integrate specialized tools and large multimodal models
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论图像编辑模型正快速进步，但全面评测仍是一项重大挑战。现有图像编辑基准普遍存在任务范围有限、评测维度不足以及严重依赖人工标注等问题，显著制约了其可扩展性与实际适用性。为解决这一问题，我们提出 I2I-Bench，一个面向图像到图像编辑模型的综合基准，其特点包括：(i) 多样化的任务，涵盖单图编辑与多图编辑两类任务下的 10 个任务类别；(ii) 全面的评测维度，包含 30 个解耦且细粒度的评测维度，并配有融合专用工具与大型多模态模型（LMM）的自动化混合评测方法；(iii) 严格的对齐验证，用以证明本基准的评测结果与人类偏好之间的一致性。利用 I2I-Bench，我们对众多主流图像编辑模型进行了基准测试，考察了各编辑模型在不同维度上的差距与权衡。我们将开源 I2I-Bench 的全部组件，以促进未来研究。
英文原文摘要 1136 字符
Image editing models are advancing rapidly, yet comprehensive evaluation remains a significant challenge. Existing image editing benchmarks generally suffer from limited task scopes, insufficient evaluation dimensions, and heavy reliance on manual annotations, which significantly constrain their scalability and practical applicability. To address this, we propose I2I-Bench, a comprehensive benchmark for image-to-image editing models, which features (i) diverse tasks, encompassing 10 task categories across both single-image and multi-image editing tasks, (ii) comprehensive evaluation dimensions, including 30 decoupled and fine-grained evaluation dimensions with automated hybrid evaluation methods that integrate specialized tools and large multimodal models (LMMs), and (iii) rigorous alignment validation, justifying the consistency between our benchmark evaluations and human preferences.Using I2I-Bench, we benchmark numerous mainstream image editing models, investigating the gaps and trade-offs between editing models across various dimensions. We will open-source all components of I2I-Bench to facilitate future research.