avatar

Zujie Liang

To the stars, through hardships.

Bio

My name is Zujie Liang (梁祖杰). I am currently a Researcher at ByteDance. Now I am working on Agentic post-training for Doubao. Before that, I obtained my M.S. degree from Sun Yat-Sen University (SYSU) in 2022. Overall, I focus on the following research topics🔬:

  • LLM/VLM Agents: Building useful LLM/VLM Agents that leverage agentic capabilities to solve real-world tasks, including Agentic Memory [A-MEM], Agentic AutoML [I-MCTS], Shopping Agents [iAgent][NTRD], Multimodal Agents [Maria] (Probably one of the earliest attempts at VLM-based agents).

  • LLM Post-training: Exploring post-training, dedicated to efficient knowledge distillation [SLMRec], alleviating the hallucination [CoKE], and enhancing the reasoning capabilities [MultiLingPoT].

  • Long Context Processing: Enhancing the long context capablities of LLMs through sparse attention design [PowerAttention] and context management framework [SEGMENT+].

  • LLM PEFT: Developing practical approaches for effectively continual training [Lottery Prompt] and multi-task learning [HPT].

  • LLM Benchmarking: Benchmarking the performance of LLMs on several KG tasks, including Concept Reasoning [CR-LLM], Historical Analogy [PmP], and Quotation Generation [QUILL].

  • Multimodal Hallucination: Focusing on alleviating the multimodal hallucination caused by modality bias. [LPF][CL-VQA]

I’m always open to discussion or collaboration. Please drop me an E-mail if you’re interested.

News

  • [08/2025] A-MEM is accepted by NeurIPS 2025, Congrats to Wujiang!
  • [08/2025] MultiLingPoT is accepted by EMNLP 2025.
  • [07/2025] Our paper on LLM Historical Analogy received an Outstanding Paper Award (Top 0.3%) in ACL 2025, Congrats to Nianqi!
  • [05/2025] 2 papers are accepted by ACL 2025.
  • [01/2025] One paper is accepted by ICLR 2025.
  • [09/2024] One paper is accepted by EMNLP 2024.
  • [05/2024] One paper is accepted by Findings of ACL 2024.
  • [08/2023] One paper is accepted by CIKM 2023.
  • [06/2023] Hosted Ant FinTech AI Challenge (AFAC2023) at TIANCHI.
  • [05/2023] One paper about Lottery Prompt Tuning is accepted by ACL 2023.
  • [07/2022] I start my journey at Ant Group.
  • [04/2022] One paper is accepted by ICMR 2022.
  • [08/2021] One paper on recommender dialogue system is accepted by EMNLP 2021.
  • [06/2021] I am offered an ACM SIGIR 2021 Student Travel Grant.
  • [06/2021] I serve as PC Member for NLPCC 2021.
  • [05/2021] One paper on visual knowledge powered conversational agent is accepted by ACL 2021.
  • [03/2021] One paper on De-biased VQA is accepted by SIGIR 2021.
  • [09/2020] One paper on Robust VQA is accepted by EMNLP 2020.

Work Experiences

  • ByteDance,
    Researcher, Aug. 2025 - Present.
  • Ant Group,
    Senior Research Engineer, July. 2022 - Aug. 2025.
  • Alibaba,
    Ads Algorithm Intern, June. 2021 - Sept. 2021.
  • Microsoft,
    Research Intern, Oct. 2020 - June. 2021.
  • Huawei,
    SDE Intern, Jun. 2018 - Sept. 2018.

Publications

LLM/VLM Agents

  • A-MEM: Agentic Memory for LLM Agents
    Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, Yongfeng Zhang
    NeurIPS 2025. [paper][code]

  • I-MCTS: Enhancing Agentic AutoML via Introspective Monte Carlo Tree Search
    Zujie Liang, Feng Wei, Wujiang Xu, Lin Chen, Yuxi Qian, Xinhui Wu
    preprint. [paper][code]

  • iAgent: LLM Agent as a Shield between User and Recommender Systems
    Wujiang Xu, Yunxiao Shi, Zujie Liang, Xuying Ning, Kai Mei, Kun Wang, Xi Zhu, Min Xu, Yongfeng Zhang
    ACL 2025 (Findings). [paper][code]

  • Maria: A Visual Experience Powered Conversational Agent
    Zujie Liang, Huang Hu, Can Xu, Chongyang Tao, Xiubo Geng, Yining Chen, Fan Liang, Daxin Jiang.
    ACL 2021. [paper][slides][code]

  • Learning Neural Templates for Recommender Dialogue System
    Zujie Liang, Huang Hu, Can Xu, Jian Miao, Yingying He, Yining Chen, Xiubo Geng, Fan Liang, Daxin Jiang
    EMNLP 2021. [paper][code]

LLM Post-training

  • SLMRec: Distilling Large Language Models into Small for Sequential Recommendation
    Wujiang Xu, Qitian Wu, Zujie Liang, Jiaojiao Han, Xuying Ning, Yunxiao Shi, Wenfang Lin, Yongfeng Zhang
    ICLR 2025. [paper][code]

  • Teaching Large Language Models to Express Knowledge Boundary from Their Own Signals
    Lida Chen, Zujie Liang, Xintao Wang, Jiaqing Liang, Yanghua Xiao, Feng Wei, Jinglei Chen, Zhenghong Hao, Bing Han, Wei Wang
    ACL 2025, KnowFM Workshop. [paper]

  • MultiLingPoT: Enhancing Mathematical Reasoning with Multilingual Program Fine-tuning
    Nianqi Li, Zujie Liang, Siyu Yuan, Jiaqing Liang, Feng Wei, Yanghua Xiao
    EMNLP 2025 (Findings). [paper][code]

Long-context LLM

  • PowerAttention: Exponentially Scaling of Receptive Fields for Effective Sparse Attention
    Lida Chen, Dong Xu, Chenxin An, Xintao Wang, Yikai Zhang, Jiangjie Chen, Zujie Liang, Feng Wei, Jiaqing Liang, Yanghua Xiao, Wei Wang
    preprint. [paper][code]

  • Segment+: Long Text Processing with Short-Context Language Models
    Wei Shi, Shuang Li, Kerun Yu, Jinglei Chen, Zujie Liang, Xinhui Wu, Yuxi Qian, Feng Wei, Bo Zheng, Jiaqing Liang, Jiangjie Chen, Yanghua Xiao
    EMNLP 2024. [paper]

LLM PEFT

  • Prompts Can Play Lottery Tickets Well: Achieving Lifelong Learning Information Extraction via Lottery Prompt Tuning
    Zujie Liang, Feng Wei, Jie Yin, Yuxi Qian, Zhenghong Hao, Bing Han.

    ACL 2023. [paper][code]

  • Hierarchical Prompt Tuning for Few-Shot Multi-Task Learning
    Jingping Liu, Tao Chen, Zujie Liang, Haiyun Jiang, Yanghua Xiao, Feng Wei, Yuxi Qian, Zhenghong Hao, Bing Han.
    CIKM 2023. [paper]

LLM Benchmarking

  • CR-LLM: A Dataset and Optimization for Concept Reasoning of Large Language Models
    Nianqi Li, Jingping Liu, Sihang Jiang, Haiyun Jiang, Yanghua Xiao, Jiaqing Liang, Zujie Liang, Feng Wei, Jinglei Chen, Zhenghong Hao, Bing Han
    ACL 2024 (Findings). [paper]

  • Past Meets Present: Creating Historical Analogy with Large Language Models
    Nianqi Li, Siyu Yuan, Jiangjie Chen, Jiaqing Liang, Feng Wei, Zujie Liang, Deqing Yang, Yanghua Xiao
    ACL 2025, oral. [paper][Outstanding Paper Award](Top 0.3%)

  • QUILL: Quotation Generation Enhancement of Large Language Models
    Jin Xiao, Bowei Zhang, Qianyu He, Jiaqing Liang, Feng Wei, Jinglei Chen, Zujie Liang, Deqing Yang, Yanghua Xiao
    preprint. [paper]

Multimodal Hallucination

  • LPF: A Language-Prior Feedback Objective Function for De-biased Visual Question Answering
    Zujie Liang, Haifeng Hu, Jiaying Zhu.
    SIGIR 2021. [paper][slides][code]

  • Learning to Contrast the Counterfactual Samples for Robust Visual Question Answering
    Zujie Liang, Weitao Jiang, Haifeng Hu, Jiaying Zhu.
    EMNLP 2020. [paper][slides][code]

Academic Services

  • Reviewer: ARR 2025, ARR 2024, ICLR 2023, EACL 2023, ACL 2022, NLPCC 2022, NLPCC 2021, IJCAI 2021

Miscellanous