kohsei/MultiBanana-Benchmark

🍌 MultiBanana: A Challenging Benchmark for Multi-Reference Text-to-Image Generation 🍌

This repository provides the datasets for

“MultiBanana: A Challenging Benchmark for Multi-Reference Text-to-Image Generation” by Yuta Oshima, Daiki Miyake, Kohsei Matsutani, Yusuke Iwasawa, Masahiro Suzuki, Yutaka Matsuo and Hiroki Furuta (2025)

Paper Link

https://arxiv.org/abs/2511.22989

Github Repository

https://github.com/matsuolab/multibanana/

Dataset Summary

Recent text-to-image generation models have acquired the ability of multi-reference generation and editing; the ability to inherit the appearance of subjects from multiple reference images and re-render them under new contexts. However, the existing benchmark datasets often focus on the generation with single or a few reference images, which prevents us from measuring the progress on how model performance advances or pointing out their weaknesses, under different multi-reference conditions. In addition, their task definitions are still vague, typically limited to axes such as "what to edit" or "how many references are given", and therefore fail to capture the intrinsic difficulty of multi-reference settings. To address this gap, we introduce MultiBanana, which is carefully designed to assesses the edge of model capabilities by widely covering multi-reference-specific problems at scale: (1) varying the number of references, (2) domain mismatch among references (e.g., photo vs. anime), (3) scale mismatch between reference and target scenes, (4) references containing rare concepts (e.g., a red banana), and (5) multilingual textual references for rendering. Our analysis among a variety of text-to-image models reveals their superior performances, typical failure modes, and areas for improvement. MultiBanana will be released as an open benchmark to push the boundaries and establish a standardized basis for fair comparison in multi-reference image generation.

Acknowledgement

This dataset partially incorporates a subset of images from the LAION-5B dataset. We acknowledge and thank the LAION team for making such a valuable large-scale dataset openly available to the research community.

Opt-out and Removal

This dataset contains images from the LAION-5B dataset, which are available on the public internet. If your image represents you or your copyrighted work and you wish to have it removed from this dataset, please contact us at {yuta.oshima, daiki.miyake}[at]g.ecc.u-tokyo.ac.jp with the relevant image. We will remove the entry from the dataset immediately.

Safety and Ethics

While this dataset is derived from LAION, web-crawled data may still contain inappropriate or harmful content. We have applied auto filtering with Gemini-Flash-2.5 and human annotator. Please exercise caution when you viewing the images.

Citation

@inproceedings{oshima2025multibanana,
  title={MultiBanana: A Challenging Benchmark for Multi-Reference Text-to-Image Generation},
  author={Yuta Oshima and Daiki Miyake and Kohsei Matsutani and Yusuke Iwasawa and Masahiro Suzuki and Yutaka Matsuo and Hiroki Furuta},
  year={2025},
  eprint={2511.22989},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2511.22989},
}