Meilong Xu

Multimodal AI · Generative Models · Topology

About Me

I am a final-year Computer Science Ph.D. Candidate at Stony Brook University, advised by Prof. Chao Chen and working closely with Dr. Xiaoling Hu. My research connects multimodal foundation models, generative modeling, and structure-aware learning. I develop methods that help models understand complex visual structures, adapt to new tasks, and produce more reliable outputs.

My recent work explores vision-language model post-training through supervised fine-tuning and reinforcement learning, alongside topology-guided diffusion models and computational pathology. At Amazon Prime Video, I work on physics-grounded video generation; previously, at TikTok, I worked on multimodal model training and video understanding. Before my Ph.D., I received my B.Eng. from Wuhan University, where I worked with Prof. Jiayi Ma on image fusion.

Research Interests

Multimodal Foundation Models & Post-Training

Adapting vision-language models for video understanding and structural reasoning through supervised fine-tuning and reinforcement learning.

Generative AI & Diffusion Models

Building structure-aware diffusion models and aligning generation with domain knowledge and human preferences.

Trustworthy & Structure-Aware Learning

Preserving meaningful topology and structural consistency while learning from limited labels and noisy predictions.

Multimodal Learning for Healthcare

Connecting histopathology with gene expression and language to learn informative, efficient representations.

Selected Publications

All publications

Preprints

arXiv preprint 2026

Topo-R1: Detecting Topological Anomalies via Vision-Language Models

Meilong Xu, Qingqiao Hu, Xiaoling Hu, Shahira Abousamra, Xin Yu, Weimin Lyu, Kehan Qi, Dimitris Samaras, Chao Chen

Equips vision-language models with topology-aware perception capabilities through supervised fine-tuning (SFT) and GRPO with topology-aware verifiable rewards.

arXiv

BibTeX

Citation for the arXiv version.

@misc{topo-r1,
  title = {Topo-R1: Detecting Topological Anomalies via Vision-Language Models},
  author = { Meilong Xu and Qingqiao Hu and Xiaoling Hu and Shahira Abousamra and Xin Yu and Weimin Lyu and Kehan Qi and Dimitris Samaras and Chao Chen },
  year = {2026},
  eprint = {2603.13054},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2603.13054}
}

arXiv preprint 2025

RB-FT: Rationale-Bootstrapped Fine-Tuning for Video Classification

Meilong Xu, Di Fu, Jiaxing Zhang, Gong Yu, Jiayu Zheng, Xiaoling Hu, Dongdi Zhao, Feiyang Li, Chao Chen, Yong Cao

Adapts vision-language models to video classification by first learning from self-generated rationales, then fine-tuning on task labels without additional annotations.

arXiv

BibTeX

Citation for the arXiv version.

@misc{rb-ft,
  title = {RB-FT: Rationale-Bootstrapped Fine-Tuning for Video Classification},
  author = { Meilong Xu and Di Fu and Jiaxing Zhang and Gong Yu and Jiayu Zheng and Xiaoling Hu and Dongdi Zhao and Feiyang Li and Chao Chen and Yong Cao },
  year = {2025},
  eprint = {2511.15923},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2511.15923}
}

Peer-reviewed Publications

2026

UPDiff-UDA framework figure from the paper

ECCV 2026Accepted

Multi-Channel Uncertainty-Weighted Score Matching for Conditional Diffusion in Medical UDA

Chen Li, Meilong Xu, Xiaoling Hu, Weimin Lyu, Chao Chen

Combines confidence-weighted conditional diffusion scores and appearance adaptation to generate training data robust to noisy pseudo-labels for cross-modality medical segmentation.

arXiv

BibTeX

Citation for the arXiv version.

@misc{updiff-uda,
  title = {Multi-Channel Uncertainty-Weighted Score Matching for Conditional Diffusion in Medical UDA},
  author = { Chen Li and Meilong Xu and Xiaoling Hu and Weimin Lyu and Chao Chen },
  year = {2025},
  eprint = {2509.22476},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2509.22476}
}

RankByGene framework figure from the paper

IEEE Transactions on Medical Imaging (TMI) 2026Accepted

RankByGene: Gene-Guided Histopathology Representation Learning Through Cross-Modal Ranking Consistency

Wentao Huang, Meilong Xu, Xiaoling Hu, Shahira Abousamra, Aniruddha Ganguly, Saarthak Kapse, Alisa Yurovsky, Prateek Prasanna, Tahsin Kurc, Joel Saltz, Michael L. Miller, Chao Chen

Aligns histology and gene-expression representations through cross-modal ranking consistency, with teacher–student self-distillation stabilizing image features.

arXiv Code Paper

BibTeX

Citation for the arXiv version.

@misc{rankbygene,
  title = {RankByGene: Gene-Guided Histopathology Representation Learning Through Cross-Modal Ranking Consistency},
  author = { Wentao Huang and Meilong Xu and Xiaoling Hu and Shahira Abousamra and Aniruddha Ganguly and Saarthak Kapse and Alisa Yurovsky and Prateek Prasanna and Tahsin Kurc and Joel Saltz and Michael L. Miller and Chao Chen },
  year = {2024},
  eprint = {2411.15076},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2411.15076}
}

Diffusion-LPO generated image examples from the paper

ICLR 2026

Towards Better Optimization For Listwise Preference in Diffusion Models

Jiamu Bai, Xin Yu, Meilong Xu, Weitao Lu, Xin Pan, Kiwan Maeng, Daniel Kifer, Jian Wang, Yu Wang

Extends diffusion preference optimization from pairs to ranked image lists, aligning generation, editing, and personalized outputs with human preferences.

arXiv Paper

BibTeX

Citation for the arXiv version.

@misc{diffusion-lpo,
  title = {Towards Better Optimization For Listwise Preference in Diffusion Models},
  author = { Jiamu Bai and Xin Yu and Meilong Xu and Weitao Lu and Xin Pan and Kiwan Maeng and Daniel Kifer and Jian Wang and Yu Wang },
  year = {2025},
  eprint = {2510.01540},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2510.01540}
}

2025

NeurIPS 2025

MATCH: Multi-faceted Adaptive Topo-Consistency for Semi-Supervised Histopathology Segmentation

Meilong Xu^*, Xiaoling Hu^*, Shahira Abousamra, Chen Li, Chao Chen

* Equal contribution

Matches persistent topological features across stochastic predictions and training snapshots to improve structural consistency in semi-supervised histopathology segmentation.

arXiv Code Paper

BibTeX

Citation for the arXiv version.

@misc{match,
  title = {MATCH: Multi-faceted Adaptive Topo-Consistency for Semi-Supervised Histopathology Segmentation},
  author = { Meilong Xu and Xiaoling Hu and Shahira Abousamra and Chen Li and Chao Chen },
  year = {2025},
  eprint = {2510.01532},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2510.01532}
}

TopoCellGen framework figure from the paper

CVPR 2025Oral

TopoCellGen: Generating Histopathology Cell Topology with a Diffusion Model

Meilong Xu, Saumya Gupta, Xiaoling Hu, Chen Li, Shahira Abousamra, Dimitris Samaras, Prateek Prasanna, Chao Chen

Integrates topological constraints into diffusion models to generate multi-class cell layouts and introduces Topological Fréchet Distance to evaluate their structure.

arXiv Code Paper Slides

BibTeX

Citation for the arXiv version.

@misc{topocellgen,
  title = {TopoCellGen: Generating Histopathology Cell Topology with a Diffusion Model},
  author = { Meilong Xu and Saumya Gupta and Xiaoling Hu and Chen Li and Shahira Abousamra and Dimitris Samaras and Prateek Prasanna and Chao Chen },
  year = {2024},
  eprint = {2412.06011},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2412.06011}
}

2024

TopoSemiSeg framework figure from the paper

ECCV 2024

Semi-supervised Segmentation of Histopathology Images with Noise-Aware Topological Consistency

Meilong Xu, Xiaoling Hu, Saumya Gupta, Shahira Abousamra, Chao Chen

Learns from unlabeled histopathology images with noise-aware topological consistency, separating reliable structural signals from noisy teacher and student predictions.

arXiv Code Slides Poster

BibTeX

Citation for the arXiv version.

@misc{toposemiseg,
  title = {Semi-supervised Segmentation of Histopathology Images with Noise-Aware Topological Consistency},
  author = { Meilong Xu and Xiaoling Hu and Saumya Gupta and Shahira Abousamra and Chao Chen },
  year = {2023},
  eprint = {2311.16447},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2311.16447}
}

Education

Stony Brook University

Ph.D. in Computer Science

Expected May 2027 · Advisor: Prof. Chao Chen

Sep 2022 – Present

Wuhan University

B.Eng. in Electronic Information Engineering

Advisor: Prof. Jiayi Ma

Sep 2018 – Jun 2022

Academic Service

Conference Reviewer

CVPR
NeurIPS
ICLR
ICML
ICCV
ECCV
ACL
EMNLP

Journal Reviewer

IEEE TPAMI
IEEE TMI
IEEE TNNLS

Awards & Honors

NeurIPS Scholar Award2025
National Scholarship, Wuhan University2019–2020
First Class Scholarship, Wuhan University2018–2021

About Me

Research Interests

Multimodal Foundation Models & Post-Training

Generative AI & Diffusion Models

Trustworthy & Structure-Aware Learning

Multimodal Learning for Healthcare

Recent News

Selected Publications

Preprints

Peer-reviewed Publications

Education

Stony Brook University

Wuhan University

Academic Service

Conference Reviewer

Journal Reviewer

Awards & Honors