Skip to content
View wangxingjun778's full-sized avatar
🎯
Focusing
🎯
Focusing

Organizations

@modelscope @embodied-agent

Block or report wangxingjun778

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
wangxingjun778/README.md

Hi there! πŸ‘‹ I'm Xingjun Wang (wangxingjun778)

Researcher/Engineer @ Alibaba Tongyi Lab, ModelScope Team
Open Source Advocate | LLM & Agent Researcher | PKU & XJTU Alumnus


πŸš€ About Me

I am a Researcher/Engineer at Alibaba Tongyi Lab, where I design and implement dataset infrastructure, LLM evaluation framework, and autonomous agent system for the ModelScope community.

As a core maintainer of ModelScope SDK, EvalScope, MS-Agent, Sirchmunk and FaceChain, my works follow the "code-first, research-driven" paradigm.

I earned my M.S. from Peking University (PKU) and my B.S. from Xi'an Jiaotong University (XJTU).


πŸŽ“ Research Interests

  • Autonomous Agent Systems: Developing Long-Horizon Agents with autonomous exploration capabilities and advancing Agentic Reinforcement Learning.
  • Unified Multimodal Models: Architecting and optimizing Any-to-Any multimodal understanding and generation systems (e.g., Nexus-Gen).
  • Multimodal Reasoning: Enhancing complex logical inference and step-by-step reasoning in cross-modal architectures.
  • Automated LLM Evaluation: Designing scalable, reliable, and bias-aware benchmarking methodologies for next-generation foundation models.

πŸ† Selected Publications

Eligen: Entity-level controlled image generation with regional attention > H. Zhang, Z. Duan, Xingjun Wang, Y. Chen, Y. Zhang.
ACM Multimedia Asia 2025 | πŸ† Best Paper Award > [Paper]

SWIFT: A Scalable lightWeight Infrastructure for Fine-Tuning > Y. Zhao, J. Huang, J. Hu, Xingjun Wang, et al.
AAAI 2025 (System Demo) | πŸ› οΈ Core Infrastructure for LLM Fine-tuning | πŸ”₯ 10k+ Stars on GitHub> [Paper] | [Code]

UniME: Breaking the Modality Barrier: Universal Embedding Learning with Multimodal LLMs > T. Gu, K. Yang, Z. Feng, Xingjun Wang, et al.
ACM Multimedia 2025

Hie-SQL: History Information Enhanced Network for Context-Dependent Text-to-SQL Semantic Parsing > Y. Zheng, H. Wang, B. Dong, Xingjun Wang, C. Li.
ACL 2022 (Findings)

Nexus-Gen: A Unified Model for Image Understanding, Generation, and Editing > H. Zhang, Z. Duan, Xingjun Wang, et al.
arXiv 2025 | [ModelScope]

FaceChain: A Playground for Identity-Preserving Portrait Generation > Y. Liu, C. Yu, L. Shang, Xingjun Wang, et al.
arXiv 2023 | πŸ”₯ 9.5k+ Stars on GitHub > [Code]

🐿️ Sirchmunk: Raw Data to Self-Evolving Intelligence, Real-Time > Xingjun Wang, ModelScope Team, et al.
Agentic Search & Embedding-Free RAG Framework | πŸ† GitHub Trending #6 Repository of The Day > [Code] | [Documentation]


βš–οΈ Professional Services

  • Program Committee Member | Reviewer: The 40th Annual AAAI Conference on Artificial Intelligence (AAAI-26)

Since Dec 30, 2025 VISITORS

Pinned Loading

  1. modelscope/modelscope modelscope/modelscope Public

    ModelScope: bring the notion of Model-as-a-Service to life.

    Python 9k 956

  2. modelscope/facechain modelscope/facechain Public

    FaceChain is a deep-learning toolchain for generating your Digital-Twin.

    Jupyter Notebook 9.5k 882

  3. modelscope/evalscope modelscope/evalscope Public

    A streamlined and customizable framework for efficient large model (LLM, VLM, AIGC) evaluation and performance benchmarking.

    Python 3k 410

  4. QwenLM/Qwen3 QwenLM/Qwen3 Public

    Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud.

    Python 27.4k 2k

  5. modelscope/ms-swift modelscope/ms-swift Public

    Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, …

    Python 14.7k 1.5k

  6. modelscope/ms-agent modelscope/ms-agent Public

    MS-Agent: a lightweight framework to empower agentic execution of complex tasks

    Python 4.3k 509