MACRO: Advancing Multi-Reference Image Generation with Structured Long-Context Data
This addresses a critical bottleneck for real-world applications like multi-subject composition and narrative illustration, representing an incremental advance through new data and evaluation tools.
The paper tackles the problem of performance degradation in multi-reference image generation as the number of input references increases, by introducing MacroData, a large-scale dataset of 400K samples with up to 10 reference images, and MacroBench, a benchmark of 4,000 samples, which together lead to substantial improvements in generative coherence.
Generating images conditioned on multiple visual references is critical for real-world applications such as multi-subject composition, narrative illustration, and novel view synthesis, yet current models suffer from severe performance degradation as the number of input references grows. We identify the root cause as a fundamental data bottleneck: existing datasets are dominated by single- or few-reference pairs and lack the structured, long-context supervision needed to learn dense inter-reference dependencies. To address this, we introduce MacroData, a large-scale dataset of 400K samples, each containing up to 10 reference images, systematically organized across four complementary dimensions -- Customization, Illustration, Spatial reasoning, and Temporal dynamics -- to provide comprehensive coverage of the multi-reference generation space. Recognizing the concurrent absence of standardized evaluation protocols, we further propose MacroBench, a benchmark of 4,000 samples that assesses generative coherence across graded task dimensions and input scales. Extensive experiments show that fine-tuning on MacroData yields substantial improvements in multi-reference generation, and ablation studies further reveal synergistic benefits of cross-task co-training and effective strategies for handling long-context complexity. The dataset and benchmark will be publicly released.