About Me

My name is Handong Zhao. I am a Senior Research Scientist in Vision Language Lab (VLL) at Adobe Research in San Jose, CA. My research focuses on vision-language foundation models, multimodal agents, and post-training, with applications in visual understanding, generation, and tool-using creative AI systems. At Adobe, I have led and co-led research efforts on multimodal image editing agents, large-scale vision-language foundation models, and image/video generation post-training, with technologies transferred across Adobe Creative Cloud products.

I received my Ph.D. in Computer Engineering from the Department of Electrical and Computer Engineering at Northeastern University , Boston, MA. I was also a recipient of the National Science Foundation Innovation Corps (I-Corps™) award as an Entrepreneurial Lead.

Recent News

  • [09/26] One paper accepted to NeurIPS 2026: LogicDirector: Enforcing Temporal Composition in Text-to-Video Generation . LogicDirector uses logic-guided test-time latent search to improve temporal composition in video generation.
  • [07/26] One paper accepted to COLM 2026: SpyRL: Task Transformation for Non-Verifiable Open-Ended Tasks . SpyRL introduces an alternative approach to solving non-verifiable tasks , distinct from our previous CVPR 2026 work RetouchIQ .
  • [05/26] Two ICML-26 papers accepted on post-training techniques, covering VividCam for Video Generation and Efficient Multimodal RL .
  • [02/26] One CVPR-26 paper accepted on RetouchIQ: MLLM Image Editing Agent via Tool-use .
  • [01/26] Two ICLR-2026 papers accepted on VLM for Controllable Retrieval and Visual Reasoning .
  • [2026] Serving as ICLR-26/ICML-26/ACL ARR-26 Area Chair.

  • More ...

    • [09/25] One NeurIPS-25 accepted on MLLM on Video Temporal Reasoning .
    • [08/25] Two EMNLP papers accepted on GUI-Agent and TableQA.
    • [06/25] Two VLM papers accepted. CVPR-25 on PhotoEye: image quality/aesthetic understanding and ICCV-25 on VSP: assessing spatial planning capability .
    • [05/25] One LLM paper accepted by ACL-25 on Augmenting Generative Decoders with Representation Learning and Infilling Capabilities.
    • [10/24] Appointed as ICLR-25 Area Chair.
    • [09/24] One NeurIPS-24 and two EMNLP-24 papers accepted on VL representation learning and MLLM .
    • [07/24] One ECCV-2024 accepted on Training-free Domain Generalization .
    • [05/24] One TMLR journal accepted on Continual Graph Learning .
    • [01/24] One ICLR-2024 accepted on Self-supervised Learning for Entity Segmentation .
    • [09/23] One NeurIPS-2023 paper accepted on InfoPrompt for NLU .
    • [07/23] One ICCV-2023 paper accepted on Spatial-Temporal Attention Controlled Diffusion Model .
    • [05/23] One ACL-2023 finding paper accepted on " Domain Adaptation for Named Entity Recognition ".
    • [03/23] One CVPR-2023 paper accepted on " Uncovering Disentaglement Capability of Diffusion Models ".
    • [01/23] Appointed as Associate Editor for for IEEE TCSVT.
    • [01/23] One ICLR-2023 paper accepted on " Continual Federated Learning ".
    • [11/22] One AAAI-2023 paper accepted on " Compositional Image Retrieval ".
    • [10/22] Two EMNLP-2022 papers accepted on " Multilingual Rep. Learning ".
    • [08/22] Four CIKM-2022 papers accepted.
    • [06/22] One RecSys-2022 on " Conversational Bundle Rec " accepted as oral.
    • [06/22] Gave a talk on Federated Knowledge Composition in CVPR FedVision Workshop. [video recording]
    • [05/22] One KDD-2022 on " KG Infusion for Tabular Pre-training Models " accepted.
    • [04/22] One SIGIR-2022 on " Cross-domain Interaction Recommendation " accepted.
    • [03/22] One CVPR-2022 on " Fashion CLIP " and one L4DC-2022 on " Neural Spatiotemporal Point Process " accepted.
    • [02/22] Two ACL-2022 papers accepted on " Few-Shot Class-Incremental NER " and " Learning Adaptive Axis Attentions in Fine-tuning (Finding paper) ".
    • [01/22] One paper " Neural Contextual Bandits with Deep Representation and Shallow Exploration " accepted by ICLR-2022 .
    • [01/22] One paper on " Federated Learning for Multilingual NLU " accepted by WWW-2022 .
    • [12/21] One paper on " Continual Federated Learning " accepted by ICLR-2023 .
    • [12/21] One paper on " Cross-Domain Clustering " accepted by AAAI-2022 student abstract and poster program.
    • [10/21] Two papers on " Continual NER " and " User-in-the-loop NER " accepted by NeurIPS ENLSP Workshop .
    • [09/21] Two papers on " KD-based Graph Similarity Computation " and " Document Representation Learning " accepted by NeurIPS-2021 .
    • [08/21] One short paper on " Sequential Recommender System " accepted by CIKM-2021 .
    • [07/21] Two papers on " Source-Free DA " and " Semi-Supervised DA " accepted by ICCV-2021 .
    • [05/21] One paper on " Explainable Column Annotation " accepted by KDD-2021 .
    • [05/21] One paper on " KG-based Commonsense Reasoning " accepted by Findings of ACL-2021 .
    • [04/21] I'm co-organizing The 3rd Workshop on Continual and Multimodal Learning for Internet of Things (CML-IOT) , at IJCAI-2021. The qualified papers are planned to invited as submissions to a journal special issue.
    • [03/21] One CVPR-2021 on " Document Representation Learning " and one NAACL-2021 on " KG Enrichment " got accepted.
    • [01/21] One ICLR-2021 on " Learning to Deceive KG " accepted.
    • [12/20] One NeurIPS-KR2ML workshop paper on " KG-based Commonsense Reasoning " accepted.
    • [11/20] One paper on " Learnable Subspace Clustering " accepted by TNNLS. [arXiv] .
    • [09/20] One NeurIPS-2020 on " Self-Supervised Relationship Probing " accepted.
    • [07/20] One ECCV-2020 on " Open Domain Image Manipulation " and one CIKM-2020 on " KG Reasoning for Recommendation " accepted.
    • [06/20] One paper on " Linear Quadratic Regulator " accepted by ICML-2020 .
    • [05/20] One paper on " Personalized Image Retrieval " accepted by KDD-2020 .
    • [04/20] One survey paper on " Representation Learning for User Modeling " was accepted by IJCAI-2020 .
    • [03/20] Our IJCAI-2020 tutorial " Robust Multi-view Visual Learning: A Knowledge Flow Perspective " was accepted (with prof. Zhengming Ding and prof. Ming Shao).
    • [02/20] One paper on " Cross-domain Document Object Detection " was accepted by CVPR-2020 .
    • [11/19] One SoCC-2019 poster and one BigData-2019 paper accepted.
    • [09/19] I gave a talk on " Multi-modal Representation Learning ", invited by Prof. Hongfu Liu at Brandeis.
    • [07/19] One ICCV-2019 paper got accepted. Congratulations to Jiuxiang.
    • [04/19] Two KDD-2019 papers accepted on the topics of " Sequential Adversarial Learning " and " Interpretable User Modeling ".
    • [02/19] My FIRST summer intern Jiuxiang's work on " Scene Graph Generation " got accepted by CVPR-2019 .
    • [02/19] Co-host (with Myra and Subrata) session " AI For System " won Top Ten Session in Adobe-TechSummit-2019. Got my first trophy at Adobe.
    • [12/18] Accepted the invitation to serve as PC member for IJCAI-2019 .
    • [11/18] Will host a mini-summit, " Analyze, Predict, and Visualize: when RNNs meet Adobe ", at Adobe-TechSummit-2019 with Sungchul Kim.
    • [10/18] I have one paper accepted by WSDM-2019 (Acceptance Rate: 16.4%).
    • [10/18] AAAI-2019 Tutorial " Deep Multi-view Data Analytics " was accepted (with Allan Ding and Hongfu Liu).
    • [05/18] Accepted the invitations to serve as PC for NIPS-2018 and SPC for AAAI-2019 .
    • [03/18] One book proposal (with Allan and Prof. Raymond Fu) is accepted by Springer.
    • [03/18] I accepted the invitation to serve as a member of the Program Committee for CVPR-2018 workshop on Analysis and Modeling of Faces and Gestures (AMFG).
    • [02/18] I accepted the invitation to serve as a member of the Program Committee for ICMLA-2018.
    • [01/18] I have one paper accepted by WACV-2018 .

    • Selected Publications

    For the complete list of publications, please refer to my Google Scholar page.

    * Project lead or primary advisor · † Equal advisor

    • RetouchIQ image retouching examples
      RetouchIQ: MLLM Agents for Instruction-Based Image Retouching with Generalist Reward
      TL;DR: Agent post-training for subjective creative tasks: introduces a generalist reward model with co-evolving policy–reward optimization to enable reinforcement learning for creative editing tasks that lack explicit, verifiable reward signals.
      Qiucheng Wu, Jing Shi, Simon Jenni, Kushal Kafle, Tianyu Wang, Shiyu Chang, Handong Zhao *
      CVPR 2026
      Paper Project Lead
    • SpyRL framework overview
      SpyRL: From RLVR to RLSVR — Task Transformation for Open-Ended LLM Self-Improvement
      TL;DR: LLM post-training via self-improvement: transforms open-ended tasks into self-verifiable learning problems, enabling reinforcement learning without external reward models or LLM judges.
      Qinsi Wang, Jing Shi, Huazheng Wang, Kun Wan, Yiran Wu, Bo Liu, Qingyun Wu, Hai Helen Li, Yiran Chen, Handong Zhao † , Wentian Zhao †
      COLM 2026
      Paper Code Co-Lead
    • VividCam: Learning Unconventional Camera Motions from Virtual Synthetic Videos
      TL;DR: Scalable synthetic data for video generation: introduces a scalable synthetic data pipeline and a disentangled training recipe that isolates camera motion learning from low-quality synthetic appearances, enabling diverse and unconventional camera controls while preserving visual realism.
      Qiucheng Wu, Handong Zhao * , Zhixin Shu, Jing Shi, Yang Zhang, Shiyu Chang
      ICML 2026
      Paper Project Page Project Lead
    • VSP benchmark: maze navigation, blocks world, collision and map planning
      VSP: Diagnosing the Dual Challenges of Perception and Reasoning in Spatial Planning Tasks for MLLMs
      TL;DR: Benchmarking visual-spatial planning: introduces a benchmark to systematically evaluate visual-spatial planning capabilities and reveal perception and reasoning limitations in multimodal LLMs.
      Qiucheng Wu, Handong Zhao † , Michael Saxon, Trung Bui, William Yang Wang, Yang Zhang, Shiyu Chang †
      ICCV 2025 Co-Lead
      Paper Code
    • PhotoEye benchmark overview
      The Photographer’s Eye: Teaching Multimodal Large Language Models to See, and Critique Like Photographers
      TL;DR: Foundation MLLM for photography and creative tasks: develops a photography-centric multimodal foundation model for understanding, evaluating, and reasoning about photographic aesthetics and creative decisions.
      Daiqing Qi, Handong Zhao * , Jing Shi, Simon Jenni, Yifei Fan, Franck Dernoncourt, Scott Cohen, Sheng Li
      CVPR 2025
      Paper Code Project Lead

    Selected Talk

    CVPR FedVision keynote: Federated Knowledge Composition — Handong Zhao

    Keynote on CVPR FedVision Workshop
    Federated Knowledge Composition
    Handong Zhao
    FedVision: International Workshop on Federated Learning for Computer Vision
    New Orleans, Louisiana. [ Recording ]

    -->