Siya Qi

Siya Qi

PhD student in Computer Science at King's College London

Siya’s Page!

Open to work — I'm on the job market! Seeking research positions (industry / academia) in NLP and LLMs starting 2026–2027. Feel free to reach out at siya.qi@kcl.ac.uk.

Welcome to my profile!

About Me

I am Siya Qi (齐思雅). I received the bachelor degree in School of Computer Science of BUPT (2016-2020), and the master degree in School of Artificial Intelligence of BUPT (2020-2023). Now I am a Ph.D. student in King’s College London (2023-Present), under the supervision of Dr. Lin Gui, Dr. Zheng Yuan, and Prof. Yulan He.

Research Interests

My research focuses on Natural Language Processing and Large Language Models, particularly in LLM reliability and real-world applications:

  • LLM Hallucination Evaluation and Mitigation: Detecting, evaluating, and mitigating hallucination in LLMs, from mixed-context hallucination assessment to attention-based contextual hallucination detection
  • Post-Training: Reinforcement learning for incentive training beyond verifiable domains, self-play / self-evolving pipelines with learnable information gain, and the side effects of reasoning-oriented training
  • Agent: LLM agents in real-world settings, e.g. a privacy-preserving, self-hosted personal agent for real-time health insights from wearable devices, as well as Theory-of-Mind reasoning about others’ beliefs
  • Mechanistic Interpretability: Analysing attention heads, neurons, and internal representations to explain and control model behaviour

Selected Publications

  1. ICML
    Detecting Contextual Hallucinations in Large Language Models with Frequency-Aware Attention
    S Qi, Y Chen, R Zhao, Q Zhu, Z Hu, W Liu, Y He, Z Yuan, L Gui
    ICML 2026
    PDF
  2. EMNLP
    HiMe: Real-Time Self-Hosted Personal Agent Platform for Health Insights with Wearable Devices
    W Liu, S Qi, L Zhang, L Tudor Car, Y He
    EMNLP 2026 (System Demonstrations)
    PDFCode
  3. ICML
    Position: Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain
    W Liu, S Qi, Y Du, Y He
    ICML 2026 (Position Paper Track)
    PDF
  4. ICLR
    When Thinking Backfires: Mechanistic Insights into Reasoning-Induced Misalignment
    H Yan, H Xu, S Qi, S Yang, Y He
    ICLR 2026
    PDFCode
  5. EMNLP
    NOVER: Incentive Training for Language Models via Verifier-Free Reinforcement Learning
    W Liu, S Qi, X Wang, C Qian, Y Du, Y He
    EMNLP 2025
    PDF
  6. ACL
    Evaluating LLMs' Assessment of Mixed-Context Hallucination Through the Lens of Summarization
    S Qi, R Cao, Y He, Z Yuan
    Findings of ACL 2025
    PDFCode
  7. ACL
    EnigmaToM: Improve LLMs' Theory-of-Mind Reasoning Capabilities with Neural Knowledge Base of Entity States
    H Xu, S Qi, J Li, Y Zhou, J Du, C Catmur, Y He
    Findings of ACL 2025
    PDF
  8. arXiv
    A Survey of Automatic Hallucination Evaluation on Natural Language Generation
    S Qi, L Gui, Y He, Z Yuan
    arXiv preprint
    PDFCode

Full publication list →

Education

  • PhD in Natural Language Processing, King’s College London, 2023-Present
  • Master’s in Intelligence Science and Technology, Beijing University of Post and Telecommunications (北京邮电大学), 2020-2023
  • Bachelor’s in Intelligence Science and Technology, Beijing University of Post and Telecommunications (北京邮电大学), 2016-2020

Experience

  • Intern, Microsoft Research Cambridge - Teams lab, 2026
  • Intern, Vivo - AI Research, 2023
  • Intern, Microsoft - Bing - SWE, 2022

Fun Facts

I love to create and maintain bullet journals (手帐). It’s a creative outlet that helps me organize my thoughts and plans while expressing my artistic side through decorative elements and layouts.

Contact

Feel free to reach out to me at siya.qi@kcl.ac.uk or connect with me on LinkedIn.