About Me
Hi! I'm Tianyu Zhan, a master's student at AI4GC Lab, Zhejiang University, advised by Shengyu Zhang and Jiwei Li. I am a recipient of the National Scholarship.
My research focuses on generative recommendation and search systems. I have published five CCF-A papers (3 as first author), with oral presentations at IJCAI 2025 and KDD 2025. Previously, I worked on edge-cloud collaborative recommendation and training efficiency for generative recommendation models. I am currently interning at the Taobao-Tmall main search team, exploring semantic identifier (SID) construction for generative search.
Research Interests
- Generative Recommendation — Training efficiency and model design for generative recommendation systems, including semantic identifier construction and token-level pruning.
- Generative Search & SID — Semantic identifier construction for generative retrieval in large-scale industrial search.
- Edge-Cloud Collaboration — Efficient co-optimization of recommendation models across edge and cloud environments, balancing latency, generalization, and resource constraints.
Education
- 2024 – present · M.S. in Software Engineering, College of Computer Science and Technology, Zhejiang University
- 2020 – 2024 · B.S. in Computer Science and Technology, Zhejiang University · GPA 4.55/5.0 (Rank 8/195)
Honors & Awards
- National Scholarship · Zhejiang University · 2025
- Five-Good Graduate Student · Zhejiang University · 2025
- Outstanding Graduate · Zhejiang University · 2024
- Provincial Government Scholarship · Zhejiang Province · 2023
- Guo Yilian Scholarship · Zhejiang University · 2022
- University First-Class Scholarship · Zhejiang University · 2021, 2022, 2023
Internships
- May 2026 – present · Taobao Main Search Team, Alibaba — Researching generative search and SID construction for large-scale e-commerce retrieval.
- Apr 2025 – Jul 2025 · Qwen Pilot Team, Tongyi Lab, Alibaba Cloud — Researched Agent tool-calling capabilities based on MCPBench, Alita, and AWorld frameworks; optimized tool descriptions to improve Agent's interface understanding; dynamically optimized tool structure and descriptions from historical call trajectories.
Selected Papers
WWW Short2026
RASTP: Representation-Aware Semantic Token Pruning for Generative Recommendation with Semantic Identifiers
arXiv2026
Semantic Trimming and Auxiliary Multi-step Prediction for Generative Recommendation
IJCAI2025
Oral
Device-Cloud Collaborative Correction for On-Device Recommendation
KDD2025
Oral
Collaboration of Large Language Models and Small Recommendation Models for Device-Cloud Recommendation
AAAI2025
MergeNet: Knowledge Migration Across Heterogeneous Models, Tasks, and Modalities
arXiv2025
Unlocking the Mysteries of OpenAI o1: A Survey of the Reasoning Abilities of Large Language Models
ACM MM2024