About / 个人简介
我目前是福州大学人工智能专业本科生,预计于 2027 年 6 月毕业。我已保研至北京航空航天大学计算机学院,预计于2027年9月入学。我的研究兴趣是让基础模型以更少的计算完成有效推理,并学习跨模态的统一表征。
在清华大学科研实习期间,我参与了置信度引导的推理链压缩、潜在空间推理控制与大小模型协同等研究;在中国科学技术大学科研实习期间,我探索了面向多模态检索的 Embedding 模型训练。同时,我也关注 AI 背后的系统问题,包括性能分析、并行计算和推理优化。
I am currently an undergraduate in Artificial Intelligence at Fuzhou University, expecting to graduate in June 2027. I have been admitted through the postgraduate recommendation pathway to the School of Computer Science and Engineering at Beihang University, with enrollment expected in September 2027. My research explores how to make foundation models reason more efficiently and learn unified representations across modalities.
I have worked on confidence-guided reasoning, latent steering and small–large model collaboration during a research internship at Tsinghua University. At the University of Science and Technology of China, I explored embedding model training for multimodal retrieval. I also focus on the systems side of AI, including profiling, parallel computing and inference optimization.
Publications / 论文
Yijia Zhu, Zhenning Shi, Junhan Shi, Xun Zhang, Yong Jiang, Qing Li
EMNLP 2026
Compressing long reasoning chains by preserving the cognitive leaps that matter.
保留关键认知跃迁,压缩冗长推理链。
Paper | PDF | Code
Zhenning Shi, Yijia Zhu, Junhan Shi, Xun Zhang, Lei Wang, Congcong Miao
EMNLP 2026
A self-distilled library of latent tools for selective, state-aware reasoning control.
从模型自身轨迹中提炼潜在工具,动态调控推理过程。
Paper | PDF | Code
Zhenning Shi, Yijia Zhu, Junhan Shi, Xun Zhang, Congcong Miao, Guorui Xie, Yong Jiang, Qing Li
Findings of EMNLP 2026
Letting small and large models share the work, one token at a time.
以 token 为粒度,让大小模型按需分工。
Paper | PDF | Code