MICON-Bench：统一多模态模型中多图上下文图像生成的基准评测与增强
MICON-Bench: Benchmarking and Enhancing Multi-Image Context Image Generation in Unified Multimodal Models
对做统一多模态模型与图像生成评测的人，这篇提供了多图上下文生成的六任务基准和 MLLM-as-verifier 的自动打分协议，可直接用于诊断跨图一致性与幻觉；DAR 是一种免训练、即插即用的推理期注意力调整手段，可迁移到任意扩散/自回归生成骨干上提升跨图连贯性。
原文依据 we present Dynamic Attention Rebalancing (DAR), a training-free, plug-and-play mechanism that dynamically adjusts attention during inference
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论统一多模态模型（UMM）近期的进展带来了显著提升的图像理解与生成能力。然而，尽管 Gemini-2.5-Flash-Image 等模型展现出对多张相关图像进行推理的新兴能力，现有基准很少涉及多图上下文生成的挑战，主要集中于文本到图像或单图编辑任务。在本工作中，我们提出 MICON-Bench，一个覆盖六项任务的综合基准，用于评估跨图组合、上下文推理与身份保持。我们进一步提出由 MLLM 驱动的「按检查点评测」（Evaluation-by-Checkpoint）框架，用于自动验证语义与视觉一致性，其中多模态大语言模型（MLLM）充当验证器。此外，我们提出动态注意力重平衡（DAR），一种免训练、即插即用的机制，在推理过程中动态调整注意力以增强连贯性并减少幻觉。在多种最优开源模型上的大量实验表明，MICON-Bench 在揭示多图推理挑战方面具有严谨性，DAR 在提升生成质量与跨图连贯性方面具有有效性。
英文原文摘要 1205 字符
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 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 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.