Hansi Zeng 曾翰偲
PhD Student · UMass Amherst, CICS · Information Retrieval & LLMs
About
I am a fifth‑year PhD student at the University of Massachusetts Amherst, advised by Prof. Hamed Zamani. My research centers on core information retrieval (IR) and agentic reinforcement learning (RL).
In core retrieval, I study optimization, scaling, and new search paradigms and modeling. We use curriculum‑learning–based knowledge distillation to enable compact dense retrievers to match larger models in CL‑DRD [SIGIR’22] (Best Short Paper). I analyze the scaling behavior of sparse and dense retrieval in decoder‑only LLMs [SIGIR’25]. I develop generative‑retrieval methods that generalize to web‑scale corpora [WWW’24], with search‑aware decoding that plans ahead [SIGIR’24], and explore LMs as semantic indexers bridging retrieval and representation learning [ICML’24]. Recently, I work on agentic RL: training LLMs with RL to reason and leverage web search for information‑seeking tasks (Search‑R1) [COLM’25].
Selected Publications
Experience
- Amazon, Rufus — Applied Scientist Intern · May–Sep 2025
- Google DeepMind — Student Research Intern · May 2024–Mar 2025
- Amazon — Applied Scientist Intern · May–Dec 2023
- Lowe’s — Data Science Intern · May–Aug 2022