---
title: OpenUni-RecA
canonical_url: "https://www.modelscope.cn/models/AI-ModelScope/OpenUni-RecA"
md_url: "https://www.modelscope.cn/models/AI-ModelScope/OpenUni-RecA.md"
repository: AI-ModelScope/OpenUni-RecA
last_updated: 2025-09-12
license: apache-2.0
pipeline_tag: any-to-any
tasks:
- any-to-any
library_name:
- pytorch
frameworks:
- pytorch
downloads: 105
stars: 0
---
# OpenUni-RecA
> OpenUni-RecA - AI-ModelScope 在 ModelScope 开源的模型。Reconstruction Alignment Improves Unified Multimodal Models
AI-ModelScope/OpenUni-RecA 是 ModelScope 魔搭社区上的any-to-any模型，采用 apache-2.0 许可。
- **Repository**: AI-ModelScope/OpenUni-RecA
- **License**: apache-2.0
- **Tasks**: any-to-any
- **Downloads**: 105
- **Stars**: 0
- **Last updated**: 2025-09-12
Source: https://www.modelscope.cn/models/AI-ModelScope/OpenUni-RecA
---
# Reconstruction Alignment Improves Unified Multimodal Models
The model was presented in the paper [Reconstruction Alignment Improves Unified Multimodal Models](https://huggingface.co/papers/2509.07295).
**Abstract:**
Unified multimodal models (UMMs) unify visual understanding and generation within a single architecture. However, conventional training relies on image-text pairs (or sequences) whose captions are typically sparse and miss fine-grained visual details--even when they use hundreds of words to describe a simple image. We introduce Reconstruction Alignment (RecA), a resource-efficient post-training method that leverages visual understanding encoder embeddings as dense "text prompts," providing rich supervision without captions. Concretely, RecA conditions a UMM on its own visual understanding embeddings and optimizes it to reconstruct the input image with a self-supervised reconstruction loss, thereby realigning understanding and generation. Despite its simplicity, RecA is broadly applicable: across autoregressive, masked-autoregressive, and diffusion-based UMMs, it consistently improves generation and editing fidelity. With only 27 GPU-hours, post-training with RecA substantially improves image generation performance on GenEval (0.73$\rightarrow$0.90) and DPGBench (80.93$\rightarrow$88.15), while also boosting editing benchmarks (ImgEdit 3.38$\rightarrow$3.75, GEdit 6.94$\rightarrow$7.25). Notably, RecA surpasses much larger open-source models and applies broadly across diverse UMM architectures, establishing it as an efficient and general post-training alignment strategy for UMMs.
## 🧠 Method
[](https://huggingface.co/papers/2509.07295)
[](https://arxiv.org/abs/2509.07295)
[](https://github.com/HorizonWind2004/reconstruction-alignment)
[](https://huggingface.co/collections/sanaka87/realign-68ad2176380355a3dcedc068)
[-fcd022?style=for-the-badge&logo=huggingface&logoColor=000)](https://huggingface.co/spaces/sanaka87/BAGEL-ReAlign)
[](https://reconstruction-alignment.github.io/)
## 🔥 News
- **2025.9.10**: BAGEL training code is released! Harmon training code will be released soon.
- **2025.9.9**: Our [finetuned weights](https://huggingface.co/collections/sanaka87/realign-68ad2176380355a3dcedc068) and [arXiv paper](https://arxiv.org/abs/2509.07295) are available! We expect to release the training code tomorrow.
## 🍭 Results
**RecA** achieves state-of-the-art performance on generation benchmarks with remarkable efficiency. Despite using only 1.5B parameters, RecA surpasses models with 7B-24B parameters, achieving GenEval **0.86** and DPGBench **87.21** without GPT-4o distillation data or reinforcement learning. RecA also improves BAGEL's editing performance significantly across all categories. Further two-stage fine-tuning with GPT-4o-Image distillation data enhances the score to **0.90** and **88.15** respectively.
We've tested RecA on various base architectures, including Show-o, OpenUni, Harmon, and BAGEL, consistently observing significant performance improvements across all models and benchmarks.
## 🏆 Model Zoo
A collection of RecA models on Hugging Face with benchmark performance:
| Model Name | Parameters | GenEval | DPGBench | ImgEdit | GEdit |
|------------|------------|---------|----------|---------|-------|\
| [BAGEL-RecA](https://huggingface.co/sanaka87/BAGEL-RecA) | 14B | 82.4 (+3.6) | 85.29 (+1.26) | 3.75 (+0.37) | 7.27 (+0.33) |\
| [Harmon-0.5B-RecA](https://huggingface.co/sanaka87/Harmon-0.5B-RecA) | 0.5B | 78.7 (+11.1) | 84.67 (+4.55) | - | - |\
| [Harmon-1.5B-RecA](https://huggingface.co/sanaka87/Harmon-1.5B-RecA) | 1.5B | 85.7 (+12.8) | 87.21 (+6.28) | - | - |\
| [Show-o-RecA](https://huggingface.co/sanaka87/Show-o-RecA) | 1.3B | 61.9 (+5.3) | 75.70 (+5.05) | - | - |\
| [Show-o-512x512-RecA](https://huggingface.co/sanaka87/Show-o-512x512-RecA) | 1.3B | 72.3 (+6.1) | 84.94 (+2.73) | - | - |\
| [Harmon-1.5B-RecA-plus](https://huggingface.co/sanaka87/Harmon-1.5B-RecA-plus) | 1.5B | 90.0 | 88.15 | - | - |\
| [OpenUni-RecA](https://huggingface.co/sanaka87/OpenUni-RecA) | 3.6B | 74.1 (+12.2) | 82.75 (+3.73) | - | - |
## ✨ Getting Started
For detailed instructions on installation, training, and evaluation, please refer to the respective repository READMEs:
- **[BAGEL Training Guide](https://github.com/HorizonWind2004/reconstruction-alignment/tree/main/BAGEL/README.md)**: Complete guide for BAGEL model training and evaluation.
- **[Benchmark Evaluation Guide](https://github.com/HorizonWind2004/reconstruction-alignment/tree/main/Benchmark/README.md)**: Multi-benchmark evaluation scripts and setup instructions.
## 🚧 TODO
- [x] Release our model weights on Hugging Face.
- [x] Release BAGEL training code.
- [ ] Release Harmon training code.
- [ ] Release Show-o and OpenUni training code.
- [ ] Further scale-up BAGEL training.
- [ ] Add support for new UMM architectures like Show-o2.
## 📮 Contact
For questions, feedback, or collaboration opportunities, feel free to reach out!
## ✍️ Citation
If you find RecA useful for your research, please consider citing:
```bibtex
@misc{xie2025reconstructionalignmentimprovesunified,
title={Reconstruction Alignment Improves Unified Multimodal Models},
author={Ji Xie and Trevor Darrell and Luke Zettlemoyer and XuDong Wang},
year={2025},
eprint={2509.07295},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2509.07295},
}
```
---
⭐ **If you find this project helpful, please consider giving it a star!** ⭐
[](https://www.star-history.com/#HorizonWind2004/reconstruction-alignment&Date)
