Full Professor, Renmin University of China
5 papers at NeurIPS 2025
This paper propose WebThinker, a deep research agent that empowers LRMs to autonomously search the web, navigate web pages, and draft research reports, all within its reasoning process.
This paper introduces an approach for training o1-like RAG models that retrieve and reason over relevant information step by step before generating the final answer.
We introduce UniGist, a unified gist token-based long context compression method without chunk-wise training, which significantly enhances long context retention and efficiency through a hardward-aligned design.
HawkBench is a human-labeled, multi-domain benchmark with 1,600 samples for evaluating RAG systems on diverse queries, revealing limits in generalizability and the need for adaptive strategies.
