Hi, I’m Boyi Deng (邓博以), a PhD candidate at the University of Science and Technology of China (USTC), affiliated with the Lab for Data Science, under the supervision of Prof. Fuli Feng and Prof. Wenjie Wang.

My research interests include mechanistic interpretability, large language models, and multilingual NLP, with a particular focus on how interpretability can be used to enhance practical model capabilities. Feel free to reach out if you’d like to discuss related topics.

🔥 News

• 2026.06:  🎉🎉 Starting my new internship at DeepSeek.
• 2026.04:  🎉🎉 We released our new technical report, Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models, which explores how sparse autoencoders can support practical workflows for analyzing, controlling, and improving LLMs.

💻 Internships

• 2026.06 - Now, DeepSeek, Beijing, China.
• 2024.08 - 2026.06, Tongyi Lab, Alibaba Group, Hangzhou, China.
  • Supervised by Dr. Basong Yang and Dr. Yu Wan.

📄 Technical Reports

• Qwen3 Technical Report, Contributor. Blog/Hugging Face
• Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models, First Author. Blog / Hugging Face

📝 Publications

• ACL 2025 Unveiling Language-Specific Features in Large Language Models via Sparse Autoencoders, Boyi Deng, Yu Wan, Baosong Yang, Yidan Zhang, Fuli Feng.
• ICLR 2026 SASFT: Sparse Autoencoder-guided Supervised Finetuning to Mitigate Unexpected Code-Switching in LLMs, Boyi Deng, Yu Wan, Baosong Yang, Fei Huang, Wenjie Wang, Fuli Feng.
• AAAI 2025 CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG, Boyi Deng, Wenjie Wang, Fengbin Zhu, Qifan Wang, Fuli Feng.
• EMNLP 2023 Attack Prompt Generation for Red Teaming and Defending Large Language Models, Boyi Deng, Wenjie Wang, Fuli Feng, Yang Deng, Qifan Wang, Xiangnan He.
• EMNLP 2025 P-mmeval: A Parallel Multilingual Multitask Benchmark for Consistent Evaluation of LLMs, Yidan Zhang, Yu Wan, Boyi Deng, Baosong Yang, Hao-Ran Wei, Fei Huang, Bowen Yu, Dayiheng Liu, Junyang Lin, Jingren Zhou.
• ACL 2026 Controllable LLM Reasoning via Sparse Autoencoder-Based Steering, Yi Fang, Wenjie Wang, Mingfeng Xue, Boyi Deng, Fengli Xu, Dayiheng Liu, Fuli Feng.

📖 Educations

• 2023.09 - 2028.07 (Expected), PhD Candidate, University of Science and Technology of China (USTC).
• 2019.09 - 2023.07, Bachelor, University of Science and Technology of China (USTC).
• 2016.09 - 2019.07, Chengdu No. 7 High School.

🎖 Honors and Awards

• 2025.10 National Scholarship