Abstract
Recent advancements in Unified Multimodal Models (UMMs) have enabled remarkable image understanding and generation capabilities. However, while models like Gemini-2.5-Flash-Image show emerging abilities to reason over multiple related images, existing benchmarks rarely address the challenges of multi-image context generation, focusing mainly on text-to-image or single-image editing tasks. In this work, we introduce \textbf{MICON-Bench}, a comprehensive benchmark covering six tasks that evaluate cross-image composition, contextual reasoning, and identity preservation. We further propose an MLLM-driven Evaluation-by-Checkpoint framework for automatic verification of semantic and visual consistency, where multimodal large language model (MLLM) serves as a verifier. Additionally, we present \textbf{Dynamic Attention Rebalancing (DAR)}, a training-free, plug-and-play mechanism that dynamically adjusts attention during inference to enhance coherence and reduce hallucinations. Extensive experiments on various state-of-the-art open-source models demonstrate both the rigor of MICON-Bench in exposing multi-image reasoning challenges and the efficacy of DAR in improving generation quality and cross-image coherence. Github: this https URL.
Abstract (translated)
最近在统一多模态模型（UMMs）方面的发展使得图像理解和生成能力显著提升。然而，尽管像Gemini-2.5-Flash-Image这样的模型展示了处理多个相关图片的能力，现有的基准测试却很少关注跨多张图片的上下文生成挑战，主要集中在文本到图像或单张图片编辑任务上。在本研究中，我们引入了\textbf{MICON-Bench}，这是一个全面的基准测试，涵盖了评估跨图象组合、上下文推理和身份保留这六个任务。此外，我们还提出了一种由多模态大型语言模型（MLLM）驱动的通过检查点进行自动验证框架——称为\textbf{MLLM-Driven Evaluation-by-Checkpoint}，用于检测语义和视觉一致性。进一步地，我们提出了一个名为\textbf{动态注意重新平衡（DAR）}的方法，这是一个无需训练且即插即用的机制，在推理过程中动态调整注意力权重以增强连贯性和减少幻觉现象。 在多种最先进的开源模型上的广泛实验表明，MICON-Bench能够严格揭示多图象推理中的挑战，并展示了DAR方法在提升生成质量和跨图像一致性方面的有效性。Github地址：this https URL.
URL
https://arxiv.org/abs/2602.19497
