Publications

Work in Progress

Here are my current research projects. I focus on finding-driven research: starting from observable model behavior, then developing targeted and concise mechanisms to improve performance and reliability. Papers will be updated upon acceptance.

• Adaptive-Thinking for Generative Recommendation
  Weihan Fei, advised by Prof. An Zhang
  In preparation for NeurIPS 2026
  – Balancing effectiveness and inference cost via selective reasoning in generative recommendation.

• QQMem: Hierarchical Query-to-Query Retrieval for Long-Context Agent Memory
  Weihan Fei, advised by Prof. Xiang Wang
  In preparation for NeurIPS 2026
  – Query-space alignment for more stable memory retrieval and grounded generation in long-context LLM agents.