Feng Xiang profile photo

Feng Xiang

冯祥

fengxiang0717@gmail.com

Research Focus
Multimodal LLMs, Agentic RL, Trustworthy & Verifiable Reasoning.

About Me

I am Feng Xiang, a master's student in Computer Science at Wuhan University and a Research Intern at Alibaba Group.

My research focuses on multimodal large language models and agentic reinforcement learning (including search and memory). Beyond building stronger model reasoning, I am particularly interested in trustworthy reasoning — enabling models to "know what they know and know what they don't", and to proactively gather clearly-sourced evidence to form verifiable reasoning chains. I believe this is essential for deploying LLM/MLLM agents in real-world, high-stakes scenarios. I am also exploring unified multimodal models that incorporate action modalities.

🤝 I am always looking for research collaborations. If you are interested in sharing GPU resources or discussing ideas, feel free to reach out!

Experience

Alibaba Group

Research Intern

Focusing on trustworthy reasoning for document intelligence.

2026.01 - Present

Wuhan University

M.S. in Computer Science and Technology, School of Computer Science

Weighted average: 92.71/100; rank: 16/207, Top 8%

2024.09 - Present

Lanzhou University

B.S. in Computer Science and Technology, School of Information Science and Engineering

Weighted average: 88.44/100; rank: 7/113, Top 6%

2020.09 - 2024.06

News

2026.06 New preprint: DocScope — benchmarking verifiable reasoning for trustworthy long-document understanding. [arXiv]
2026.05 Omni-I2C accepted at ACL 2026. [arXiv]
2026.01 AnesSuite accepted at ICLR 2026. [arXiv]
2026.01 Joined Alibaba Group as a Research Intern.
2025.10 New preprint: Adaptive Decoding via hierarchical neural information gradients. [arXiv]
2025.09 Received First-class Academic Scholarship, Wuhan University.
2025.08 New preprint: REX-RAG — reasoning exploration with policy correction in RAG. [arXiv]
2024.10 Received Second-class Freshman Scholarship, Wuhan University.
2024.09 National Third Prize, China Graduate Mathematical Modeling Competition.
2024.09 Started M.S. at Wuhan University.
2024.06 Outstanding Undergraduate Thesis, Lanzhou University.
2023.09 Outstanding Student Second-class Scholarship, Lanzhou University.
2023.06 National Third Prize, Lanqiao Cup Python Programming, Group A.
2022.09 Outstanding Student Second-class Scholarship, Lanzhou University.
2022.09 National Second Prize, Contemporary Undergraduate Mathematical Contest in Modeling.

Publications

  1. DocScope: Benchmarking Verifiable Reasoning for Trustworthy Long-Document Understanding
    X. Feng, J. Zhou, Z. Huang, K. Wang, S. Ye, J. Hu, Z. Chen, Y. Luo, J. Zhang
    arXiv preprint arXiv:2605.08888, 2026
    A benchmark that evaluates whether MLLMs can produce trustworthy, verifiable reasoning traces over long, visually rich documents via a four-stage evaluation protocol.
    Paper | Code
  2. Omni-I2C: A Holistic Benchmark for High-Fidelity Image-to-Code Generation
    J. Zhou, C. Zhang, X. Feng, Q. Zhang, H. Qiu, L. He, D. Ye, X. Gao, J. Zhang
    ACL 2026
    A comprehensive benchmark of 1,080 samples across 5 code types and 45 figure types for evaluating LMMs on converting complex digital graphics into executable code.
    Paper | Code
  3. AnesSuite: A Comprehensive Benchmark and Dataset Suite for Anesthesiology Reasoning in LLMs
    X. Feng, W. Jiang, Z. Wang, Y. Luo, P. Xu, B. Yu, H. Jin, J. Zhang
    International Conference on Learning Representations (ICLR), 2026
    The first comprehensive dataset suite for anesthesiology reasoning, covering benchmark, training data (CPT/SFT/RLVR), and Morpheus baseline reasoning models.
    Paper | Code
  4. REX-RAG: Reasoning Exploration with Policy Correction in Retrieval-Augmented Generation
    W. Jiang, X. Feng, Z. Wang, Y. Luo, P. Xu, Z. Chen, B. Du, J. Zhang
    arXiv preprint arXiv:2508.08149, 2025
    Addresses dead-end exploration in RL-trained RAG agents through mixed sampling with exploratory prompts and a policy correction mechanism to reduce distribution shift.
    Paper | Code
  5. Adaptive Decoding via Hierarchical Neural Information Gradients in Mouse Visual Tasks
    J. Feng, X. Feng
    arXiv preprint arXiv:2510.09451, 2025
    Proposes a hierarchical neural-information gradient framework to decode visual task representations from mouse brain activity across cortical regions.
    Paper
  6. Decoding Mouse Visual Tasks via Hierarchical Neural-Information Gradients
    J. Feng, X. Feng, Y. Luo, J. Li
    Mathematics 14(1), 31, 2025
    Studies hierarchical information gradients across mouse visual cortex to understand how neural data flows support visual task decoding.
  7. Orthogonal-moment-based Attraction Measurement with Ocular Hints in Video-watching Task
    M. Yang, X. Feng, R. Ma, X. Li, C. Mao
    IEEE Transactions on Computational Social Systems 10(3), 900-909, 2023
    Combines orthogonal moments with eye-tracking signals to measure viewer attraction levels during video-watching tasks.

Academic Service

Reviewer: Annual Conference on Neural Information Processing Systems (NeurIPS), IEEE Transactions on Multimedia (TMM), IEEE International Conference on Multimedia and Expo (ICME).
Memberships: IEEE Student Member, IEEE Geoscience and Remote Sensing Society (GRSS) Member.

Skills

Programming and ML: Python, Verl, LLaMA-Factory, LangChain, model fine-tuning, RAG pipelines.
Research tooling: LaTeX, academic writing, experiment design, Adobe Illustrator, Word, PowerPoint.