Shengjia (Cynthia) Zhang

I am a Master student at Zhejiang University in ZLST Lab, supervised by Prof. Jiawei Chen.

Email  /  GitHub  /  Google Scholar  / 

profile photo

Research (* denotes equal contribution)

I'm interested in machine learning, large language model.

project image

Talos: Optimizing Top-K Accuracy in Reommender Systems


Shengjia Zhang*, Weiqin Yang*, Jiawei Chen, Peng Wu, Yuegang Sun, Gang Wang, Qihao Shi, Can Wang
WWW 2026, 2026
arxiv / code /

Unfortunately, this work is rejected by NeurIPS 2025 with high scores 5545 (5 denotes accept, 4 denotes weak accept, the maximum score is 6).

We propose Talos, a recommendation loss which optimizes Top-K accuracy in recommender systems. Talos leverages a quantile technique, which aviods the complex ranking comparison, to smoothly optimizes Top-K accuracy.
project image

OThink-R1: Intrinsic Fast/Slow Thinking Mode Switching for Over-Reasoning Mitigation


Shengjia Zhang*, Junjie Wu*, Jiawei Chen, Changwang Zhang, Zhe Li, Xingyu Lou, Wangchunshu Zhou, Sheng Zhou, Can Wang, Jun Wang
Arxiv 2025, 2026
arxiv / media / code /

This work is reported by MIT Technology Review. See Media for more details.

We propose OThink-R1, a hybrid reasoning framework that integrates both fast-thinking and slow-thinking modes within a single large reasoning model and enables automatic mode switching based on problem characteristics.
project image

Breaking the Top-K Barrier: Advancing Top-K Ranking Metrics Optimization in Recommender Systems


Weiqin Yang, Jiawei Chen, Shengjia Zhang, Peng Wu, Yuegang Sun, Yan Feng, Chun Chen, Can Wang
KDD 2025, 2025
arxiv / code /

We propose a SL@K, a surrogate loss for optimizing NDCG@K in recommender systems.
project image

Advancing Loss Functions in Recommender Systems: A Comparative Study with a Rényi Divergence-Based Solution


Shengjia Zhang, Jiawei Chen, Changdong Li, Sheng Zhou, Qihao Shi, Yan Feng, Chun Chen, Can Wang
AAAI 2025, 2025
arxiv / code /

We propose a new robust recommendation loss function DrRL, tailored for recommender systems.

Awards

2025 National Scholarship, Zhejiang University
Ministry of Education of the People's Republic of China

2022 Outstanding Winner & Frank Giordano Award (1/15105), Mathematical Contest in Modeling (MCM)
[Certificate] [Official Results]
Consortium for Mathematics and Its Applications (COMAP)


Design and source code from Jon Barron's website