| layout | fancy_home |
|---|---|
| permalink | /publications/ |
| title | Publications |
Research Snapshot
My research lies at the intersection of Generative Models and Reward Modeling, Autonomous Agents, and Visual Perception. I build generalized frameworks like PosterCraft and PosterOmni to redefine artistic creation, design reward models for human-aligned generation PosterReward, and empower agents with reasoning capabilities for open-world understanding.
<div class="highlight-list">
<article class="highlight-card">
<div class="highlight-media">
<img src="https://ephemeral182.github.io/images/posteromni/teaser_0209.jpg" alt="PosterOmni teaser">
<span class="badge badge-cyan">CVPR 2026</span>
</div>
<div class="highlight-content">
<h3>PosterOmni — Generalized Artistic Poster Creation via Task Distillation and Unified Reward Feedback</h3>
<p class="authors"><strong>Sixiang Chen*</strong>, Jianyu Lai*, Jialin Gao*, Hengyu Shi*, Zhongying Liu*, Tian Ye, Junfeng Luo, Xiaoming Wei, Lei Zhu✉️</p>
<p class="summary">One model for poster creation—unifying local edits and global design for generalized multi-task image/poster-to-poster generation. ✨ Your intelligent assistant for high-quality aesthetic poster creation!</p>
<div class="resource-list">
<a href="https://arxiv.org/abs/2602.12127">📄 PDF</a>
<a href="https://ephemeral182.github.io/PosterOmni/">🌐 Project</a>
<a href="https://huggingface.co/MeiGen-AI/PosterOmni_v1">🤗 Model</a>
<a href="https://github.com/MeiGen-AI/PosterOmni" target="_blank" rel="noopener">🐙 GitHub</a>
<span class="gh-btn-wrap">
<iframe src="https://ghbtns.com/github-btn.html?user=MeiGen-AI&repo=PosterOmni&type=star&count=true&size=small" frameborder="0" scrolling="0" width="100" height="20" title="PosterOmni GitHub Star"></iframe>
</span>
<span class="gh-btn-wrap">
<iframe src="https://ghbtns.com/github-btn.html?user=MeiGen-AI&repo=PosterOmni&type=fork&count=true&size=small" frameborder="0" scrolling="0" width="98" height="20" title="PosterOmni GitHub Fork"></iframe>
</span>
</div>
</div>
</article>
<article class="highlight-card">
<div class="highlight-media">
<img src="https://ephemeral182.github.io/images/posterreward/teaser2.png" alt="PosterReward">
<span class="badge badge-cyan">CVPR 2026</span>
</div>
<div class="highlight-content">
<h3>PosterReward: Unlocking Accurate Evaluation for High-Quality Graphic Design Generation</h3>
<p class="authors">Jianyu Lai*, <strong>Sixiang Chen*</strong>, Jialin Gao*, Hengyu Shi, Zhongying Liu, Fuxiang Zhai, Junfeng Luo, Xiaoming Wei, Lujia Wang, Lei Zhu✉️</p>
<p class="summary">A comprehensive reward model for design aesthetics and typography, trained on a automated preference dataset to unlock accurate evaluation for high-quality graphic design generation.</p>
<div class="resource-list">
<a href="https://alexlai2860.github.io/PosterReward/">🌐 Project</a>
<a href="https://github.com/MeiGen-AI/PosterReward" target="_blank" rel="noopener">🐙 GitHub</a>
</div>
</div>
</article>
<article class="highlight-card">
<div class="highlight-media">
<img src="https://ephemeral182.github.io/images/postercraft/pub/fig1.png" alt="PosterCraft">
<span class="badge badge-green">ICLR 2026</span>
</div>
<div class="highlight-content">
<h3>PosterCraft — Rethinking High-Quality Aesthetic Poster Generation in a Unified Framework</h3>
<p class="authors"><strong>Sixiang Chen*</strong>, Jianyu Lai*, Jialin Gao*, Tian Ye, Haoyu Chen, Hengyu Shi, Shitong Shao, Yunlong Lin, Song Fei, Zhaohu Xing, Yeying Jin, Junfeng Luo, Xiaoming Wei, Lei Zhu✉️</p>
<p class="summary">A new framework for "Rethinking High-Quality Aesthetic Poster Generation in a Unified Framework". ✨ From your prompts to high-quality aesthetic posters!</p>
<div class="resource-list">
<a href="https://arxiv.org/abs/2506.10741">📄 PDF</a>
<a href="https://ephemeral182.github.io/PosterCraft/">🌐 Project</a>
<a href="https://huggingface.co/spaces/Ephemeral182/PosterCraft">🚀 Demo</a>
<a href="https://huggingface.co/PosterCraft/PosterCraft-v1_RL">🤗 Model</a>
<a href="https://huggingface.co/PosterCraft">🤗 Dataset</a>
<a href="https://github.com/Ephemeral182/PosterCraft" target="_blank" rel="noopener">🐙 GitHub</a>
<span class="gh-btn-wrap">
<iframe src="https://ghbtns.com/github-btn.html?user=Ephemeral182&repo=PosterCraft&type=star&count=true&size=small" frameborder="0" scrolling="0" width="100" height="20" title="PosterCraft GitHub Star"></iframe>
</span>
<span class="gh-btn-wrap">
<iframe src="https://ghbtns.com/github-btn.html?user=Ephemeral182&repo=PosterCraft&type=fork&count=true&size=small" frameborder="0" scrolling="0" width="98" height="20" title="PosterCraft GitHub Fork"></iframe>
</span>
</div>
</div>
</article>
<article class="highlight-card">
<div class="highlight-media">
<img src="https://ephemeral182.github.io/images/cvpr25_jarvisir.png" alt="JarvisIR">
<span class="badge badge-cyan">CVPR 2025</span>
</div>
<div class="highlight-content">
<h3>JarvisIR — Elevating Autonomous Driving Perception with Intelligent Image Restoration</h3>
<p class="authors">Yunlong Lin*, Zixu Lin*, Haoyu Chen*, Panwang Pan*, Chenxin Li, <strong>Sixiang Chen</strong>, Kairun Wen, Yeying Jin, Wenbo Li, Xinghao Ding✉️</p>
<p class="summary">A multimodal agent reasons about adverse scenes, calling specialized restoration experts to stabilize perception stacks for autonomous vehicles.</p>
<div class="resource-list">
<a href="https://arxiv.org/abs/2504.04158">📄 PDF</a>
<a href="https://cvpr2025-jarvisir.github.io/">🌐 Project</a>
<a href="https://huggingface.co/spaces/LYL1015/JarvisIR">🚀 Demo</a>
<a href="https://huggingface.co/LYL1015/JarvisIR">🤗 Model</a>
<a href="https://github.com/LYL1015/JarvisIR" target="_blank" rel="noopener">🐙 GitHub</a>
</div>
</div>
</article>
</div>
[CVPR 2026] PosterOmni — Generalized Artistic Poster Creation via Task Distillation and Unified Reward Feedback
Sixiang Chen*, Jianyu Lai*, Jialin Gao*, Hengyu Shi*, Zhongying Liu*, Tian Ye, Junfeng Luo, Xiaoming Wei, Lei Zhu
One model for poster creation—unifying local edits and global design for generalized multi-task image/poster-to-poster generation. ✨ Your intelligent assistant for high-quality aesthetic poster creation!
Jianyu Lai*, Sixiang Chen*, Jialin Gao*, Hengyu Shi, Zhongying Liu, Fuxiang Zhai, Junfeng Luo, Xiaoming Wei, Lujia Wang, Lei Zhu
A comprehensive reward model for design aesthetics and typography, trained on automated preference dataset to unlock accurate evaluation for high-quality graphic design generation.
[ICLR 2026] PosterCraft — Rethinking High-Quality Aesthetic Poster Generation in a Unified Framework
Sixiang Chen*, Jianyu Lai*, Jialin Gao*, Tian Ye, Haoyu Chen, Hengyu Shi, Shitong Shao, Yunlong Lin, Song Fei, Zhaohu Xing, Yeying Jin, Junfeng Luo, Xiaoming Wei, Lei Zhu
A new framework for "Rethinking High-Quality Aesthetic Poster Generation in a Unified Framework". ✨ From your prompts to high-quality aesthetic posters!
Sixiang Chen, Tian Ye, Yunlong Lin, Yeying Jin, Yijun Yang, Haoyu Chen, Jianyu Lai, Song Fei, Zhaohu Xing, Fugee Tsung, Lei Zhu
Proposes a one-step, reference-controllable haze generator that better matches real-world haze complexity than classic physics pipelines, creating high-quality paired data to boost real-world dehazing performance.
Sixiang Chen*, Jinbin Bai*, Zhuoran Zhao*, Tian Ye*, Qingyu Shi, Donghao Zhou, Wenhao Chai, Xin Lin, Jianzong Wu, Chao Tang, Shilin Xu, Tao Zhang, Haobo Yuan, Yikang Zhou, Wei Chow, Linfeng Li, Xiangtai Li, Lei Zhu, Lu Qi
Provides a systematic benchmark study of GPT-4o image generation across 20+ tasks (text-to-image, image-to-image, image-to-3D, image-to-X), summarizing strengths/limitations and what they imply for unified multimodal generation.
Yunlong Lin*, Zixu Lin*, Haoyu Chen*, Panwang Pan*, Chenxin Li, Sixiang Chen, Kairun Wen, Yeying Jin, Wenbo Li, Xinghao Ding
A multimodal agent reasons about adverse scenes, calling specialized restoration experts to stabilize perception stacks for autonomous vehicles.
[CVPR 2025] SnowMaster — Comprehensive Real-world Image Desnowing via MLLM with Multi-Model Feedback Optimization
Jianyu Lai*, Sixiang Chen*, Yunlong Lin, Tian Ye, Yun Liu, Song Fei, Zhaohu Xing, Hongtao Wu, Wei Wang, Lei Zhu
Uses an MMPO/DPO-enhanced MLLM as an evaluator to rank/filter pseudo-labels, enabling semi-supervised training that improves real-world desnowing without relying on dense paired GT.
Zhaohu Xing, Lihao Liu, Yijun Yang, Hongqiu Wang, Tian Ye, Sixiang Chen, Wenxue Li, Guang Liu, Lei Zhu
Builds an iterative data engine to harvest large-scale unlabeled data and progressively select high-reliability pseudo labels, improving mirror detection robustness and generalization in diverse scenes.
Tian Ye, Sixiang Chen, Haoyu Chen, Wenhao Chai, Jingjing Ren, Zhaohu Xing, Wenxue Li, Lei Zhu
Introduces a depth-prompting paradigm: leverages Depth Anything’s stable depth priors as prompts to guide dehazing, aiming for plug-and-play real-world dehazing under complex haze.
[AAAI 2025] AGLLDiff — Guiding Diffusion Models Towards Unsupervised Training-Free Real-World Low-Light Image Enhancement
Yunlong Lin*, Tian Ye*, Sixiang Chen*, Zhenqi Fu, Yingying Wang, Wenhao Chai, Zhaohu Xing, Lei Zhu, Xinghao Ding
Proposes a training-free, unsupervised diffusion guidance framework that steers a pretrained diffusion model using attribute-based guidance for effective real-world low-light enhancement.
[NeurIPS 2024] RestoreAgent — Autonomous Image Restoration Agent via Multimodal Large Language Models
Haoyu Chen, Wenbo Li, Jinjin Gu, Jingjing Ren, Sixiang Chen, Tian Ye, Renjing Pei, Kaiwen Zhou, Fenglong Song, Lei Zhu
Proposes an MLLM-driven agent that diagnoses degradations and plans a task sequence + selects expert models from a tool/model pool to restore images.
[ECCV 2024] Teaching Tailored to Talent — Adverse Weather Restoration via Prompt Pool and Depth-Anything Constraint
Sixiang Chen, Tian Ye, Kai Zhang, Zhaohu Xing, Yunlong Lin, Lei Zhu
Using a prompt pool to compose weather-specific prompts on the fly, plus Depth-Anything–constrained scene prompts to stabilize background reconstruction under unseen adverse-weather combinations.
Tian Ye, Sixiang Chen, Wenhao Chai, Zhaohu Xing, Jing Qin, Ge Lin, Lei Zhu
Presents DTPM (Diffusion Texture Prior Model) that explicitly models high-quality texture priors to preserve fine details and structure when applying diffusion to restoration.
Tian Ye*, Sixiang Chen*, Jinbin Bai*, Shi Jun, Chenghao Xue, Jingjia Jiang, Junjie Yin, Erkang Chen, Yun Liu
Formulates adverse weather removal as matching and fusing degraded features with high-quality priors stored in a learned codebook.
[ICCV 2023] Sparse Sampling Transformer with Uncertainty-Driven Ranking for Unified Removal of Raindrops and Rain Streaks
Sixiang Chen*, Tian Ye*, Jinbin Bai, Jun Shi, Erkang Chen, Lei Zhu
A transformer that uses sparse sampling attention to model global rain degradation relations and an uncertainty-driven ranking strategy to focus on hard-to-restore regions for unified deraining.
[ACM MM 2023] Uncertainty-Driven Dynamic Degradation Perceiving and Background Modeling for Efficient Single Image Desnowing
Sixiang Chen, Tian Ye, Chenghao Xue, Haoyu Chen, Yun Liu, Erkang Chen, Lei Zhu
Proposes an efficient single image desnowing framework by modeling uncertainty in degradation perception and background reconstruction.
Tian Ye*, Yunchen Zhang*, Mingchao Jiang*, Liang Chen, Yun Liu, Sixiang Chen, Erkang Chen
Argues that dehazing hinges on haze density perception, introducing Separable Hybrid Attention and density modeling to handle uneven haze distribution.