Research

Publications

My recent work focuses on image generation, unified understanding-generation models, latent visual reasoning, reward modeling, and open-world visual understanding.

2026

First AuthorUnder Review

Latent Action Control for Reasoning-Guided Unified Image Generation

Fuxiang Zhai, Sixiang Chen, Yingjin Li, Shuaibo Li, Jianyu Lai, Tengjun Huang, Lei Zhu

A reasoning-guided unified generation framework that turns planning, internal drafting, diagnosis, and refinement into hidden continuous actions consumed by the generator.

Paper
Under Review

GenEvolve: Self-Evolving Image Generation Agents via Tool-Orchestrated Visual Experience Distillation

Sixiang Chen, Zhaohu Xing, Tian Ye, Xinyu Geng, Yunlong Lin, Jianyu Lai, Xuanhua He, Fuxiang Zhai, Jialin Gao, Lei Zhu

A self-evolving image-generation agent that distills structured experience from tool-orchestrated generation trajectories.

Paper
Under Review

ROAD: Rule-Grounded Context-Aware Open-World Driver Anomaly Detection

Co-author: Fuxiang Zhai

Rule-grounded and context-aware modeling for open-world driver anomaly detection.

OpenReview
CVPR 2026 ยท CCF A

PosterReward: Unlocking Accurate Evaluation for High-Quality Graphic Design Generation

Jianyu Lai*, Sixiang Chen*, Jialin Gao*, Hengyu Shi, Zhongying Liu, Fuxiang Zhai, Junfeng Luo, Xiaoming Wei, Lujia Wang, Lei Zhu

A reward modeling framework for typography-, layout-, and aesthetics-aware evaluation of graphic design generation.

Paper Project Code