fMRI · Information Theory · Agent Memory
Three labs, three angles on one question: how do information systems hold and retrieve what matters?
AlphaLab @ USTC · Prof. Xiang Wang · First author
Path-aware auditable memory for LLM agents. State frames, evidence ledgers, and an auditor-gater protocol that lets you trace why an agent did what it did.
Under double-blind review (NeurIPS 2026)
MIRA Lab @ USTC · Prof. Jie Wang · Co-author
Cost-regularized information bottleneck for test-time context construction. How much context is enough, and what should we throw away?
Under double-blind review (NeurIPS 2026)
USTC Medical Imaging Center · Prof. Xiaoxiao Wang · Research student
Decoding neural signals during cognitive tasks. The starting point of my interest in how information is compressed and retrieved.
Completed