§ Homepage — Hong Kong · 2026

Sixiang Chen 陈思翔

Ph.D. Student Tencent Hunyuan · 青云计划 Multimodal Agents · Multimodal Coding · Computer Use · Reward & Verification.

I'm a Ph.D. student at HKUST@ROAS, advised by Prof. Lei Zhu and Prof. Fugee Tsung. My research focuses on Multimodal Agents and Generative Models, with an emphasis on Multimodal Coding, Computer Use, and Verifiable Feedback.

My industry research experience includes Tencent Hunyuan through the 青云计划 (Project Up) and Meituan LongCat (Meigen-AI) through the 北斗计划 (Beidou). My work spans multimodal coding, rubric-based reward modeling, and visual generation. I have contributed to projects including HunyuanImage 3.5, GenEvolve, JarvisHub, DeepSearch-World, Trinity, and PosterCraft. Across these efforts, I aim to connect perception, reasoning, coding, and computer use, while exploring physical AI to extend agent capabilities from digital environments into the physical world.

01 § Industry

Industry Experience

Research on multimodal agents, generative models, and verifiable feedback, spanning agent learning, tool use, and post-training.

Research Internship May 2026 — Nov 2026

Tencent Hunyuan 青云计划 · Project Up

Research Intern · Hunyuan Group

Selected as a Project Up (青云计划) research intern at Tencent Hunyuan, contributing to multimodal agents and visual foundation models, with a focus on post-training, reward modeling, and verification for learning from task outcomes and interaction.

Post-training RL · Reward Multimodal Agents Visual Generation
Previously Apr 2025 — May 2026

Meituan LongCat (Meigen-AI) 北斗计划 · Beidou

Research Intern · Vision & Generation Group

Selected as a Beidou (北斗计划) research intern, contributing to self-evolving visual agents, poster creation, and evaluation feedback. Projects include GenEvolve, PosterCraft, and PosterReward, with work released through MeiGen-AI.

Self-Evolving Agents Poster Creation Reward Modeling

02 § Research

Research Interests

I study how multimodal agents execute tasks, verify outcomes, and learn from experience — through code and software tools, structured evaluation, self-distillation, and persistent project state.

Multimodal Agents

Studying agents that learn from interaction and sustain long-horizon work. GenEvolve and DeepSearch-World explore learning from tool-use trajectories; JarvisHub provides a shared canvas and persistent project state for planning, revision, and failure recovery.

GenEvolve NeurIPS'26 JarvisHub DeepSearch-World EMNLP'26 JarvisIR CVPR'25 RestoreAgent NeurIPS'24

Multimodal Coding & Computer-use Agents

Connecting multimodal instructions to executable code and software actions. Trinity explores code-driven graphic design, while JarvisVid studies video storytelling through closed-loop planning. My broader interests include agents that operate interfaces and check the results of their actions.

Trinity JarvisVid Computer Use

Reward Modeling & Verification

Turning task requirements into checkable criteria and evidence-grounded feedback. My interests include fine-grained rubrics, multimodal judgers, reward models, and verifiers that help identify failures and guide an agent's next revision.

HunyuanImage 3.5 Preview PosterReward CVPR'26 SnowMaster CVPR'25

Visual Generation

Building visual models that generate, compose, and edit — from general-purpose image synthesis to poster creation, with an emphasis on visual quality, typography, layout, and alignment with human preferences.

HunyuanImage 3.5 Preview PosterCraft ICLR'26 PosterOmni CVPR'26 PosterReward CVPR'26

03 § News

News & Highlights

  • Sept 2026PaperGenEvolve accepted by NeurIPS 2026.
  • Sept 2026ReleaseHunyuanImage 3.5 (Preview) released, supporting text-to-image generation, reference-based generation, and multi-turn editing.
  • May 2026ReleaseReleased GenEvolve, a multi-modal agent that orchestrates search, visual references, and generation tools — and self-evolves through visual experience distillation. Co-developed at Meituan LongCat (Meigen-AI).
  • Feb 2026PaperPosterOmni accepted by CVPR 2026.
  • Feb 2026PaperPosterReward accepted by CVPR 2026.
  • Feb 2026ReleaseReleased PosterOmni — one model for poster creation, unifying local edits and global design for multi-task image/poster-to-poster generation. Your intelligent assistant for high-quality aesthetic poster creation.
  • Jan 2026PaperPosterCraft accepted by ICLR 2026.
  • Jun 2025ReleaseReleased PosterCraft, a framework for high-quality aesthetic poster generation. From your prompts to high-quality aesthetic posters.
  • Jun 2025PaperGenHaze accepted by ICCV 2025.
  • Apr 2025ReportOur comprehensive GPT-4o Image Generation Capabilities evaluation report is now available on arXiv.
  • Feb 2025PaperThree papers accepted by CVPR 2025.
  • Feb 2025RoleServing as a reviewer for NeurIPS 2025 and ICCV 2025.
  • Jan 2025RoleServing as a reviewer for IEEE TIP.
  • Dec 2024RoleServing as a reviewer for CVPR 2025 and ICML 2025.
  • Dec 2024PaperFour papers accepted by AAAI 2025.
  • Sept 2024PaperOne paper accepted by NeurIPS 2024.
  • Aug 2024RoleServing as a reviewer for ICLR 2025.
  • Jul 2024PaperTwo papers accepted by ECCV 2024.
  • Jun 2024PaperOne paper accepted by MICCAI 2024.
  • Jun 2024RoleServing as a reviewer for IEEE TITS and TCSVT.
  • May 2024RoleServing as a reviewer for NeurIPS 2024.
  • Apr 2024PaperOne paper accepted by CVPR 2024 as a Highlight.
  • Nov 2023PaperOne paper accepted by AAAI 2024.
  • Aug 2023PaperOne paper accepted by BMVC 2023.
  • Jul 2023PaperFour papers accepted by ACM MM 2023; two papers accepted by ICCV 2023.
  • Jul 2023RoleServing as a reviewer for IEEE TMM.
  • Jun 2023HonorAwarded Outstanding Graduate from Jimei University.
  • Feb 2023PaperTwo papers accepted by ICASSP 2023.
  • Sept 2022PaperOne paper accepted by ACCV 2022.
  • Jul 2022PaperOne paper accepted as an oral presentation by ECCV 2022.
  • Jun 2022PaperOne paper accepted by CVPRW 2022.

04 § Service & Awards

Academic Service & Awards

Program Committee · Reviewer

Conference on Computer Vision and Pattern Recognition (CVPR 2026)

International Conference on Learning Representations (ICLR 2026)

IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

International Conference on Computer Vision (ICCV 2025)

Conference on Neural Information Processing Systems (NeurIPS 2025)

IEEE Transactions on Image Processing (TIP)

International Conference on Machine Learning (ICML 2025)

Conference on Computer Vision and Pattern Recognition (CVPR 2025)

International Conference on Learning Representations (ICLR 2025)

IEEE Transactions on Intelligent Transportation Systems (TITS)

Conference on Neural Information Processing Systems (NeurIPS 2024)

European Conference on Computer Vision (ECCV 2024)

IEEE Transactions on Circuits and Systems for Video Technology (TCSVT)

IEEE Transactions on Multimedia (TMM)

ACM Multimedia (ACM MM 2023 / 2024)

International Conference on Computer Vision (ICCV 2023)

International Conference on Robotics and Automation (ICRA 2023)

Asian Conference on Computer Vision (ACCV 2022)

Selected Awards

CVPR Workshop · 2025
CVPR Workshop · 2024
PG Scholarship of HKUST (Guangzhou)
2024
Outstanding Graduate, Jimei University
2023
International Third Prize (Team Leader), Mathematical Contest in Modeling
2022
National Second Prize (Team Leader), China Undergraduate Mathematical Contest in Modeling
2021
National Second Prize (Team Leader), Mathorcup Mathematical Contest in Modeling
2021

05 § Education

Education & Experience

  • Sept 2024 — PresentPh.D. Student, The Hong Kong University of Science and Technology (Guangzhou)
  • Aug 2023 — Aug 2024Research Assistant, The Hong Kong University of Science and Technology (Guangzhou)
  • Sept 2019 — Jun 2023B.Eng., Jimei University