Projects

Neural Value Alignment
MSRA-IITP collaborative project on neuroscience-grounded human-AI value alignment. Drove the core research component (Meta-PCN, ICLR 2026).
Genie Brain: Brain-like Abstraction and Reasoning Engine
Devised core algorithms for brain-inspired abstraction (System 1) and reasoning (System 2). Spearheaded collaborative research on language modeling.
Optimal Frame RTB Algorithm Using Deep Reinforcement Learning
Formulated real-time bidding optimization as an MDP and designed an offline deep RL method based on counterfactual learning.
Brain-inspired AI with Human-like Intelligence
Research on cognitive brain development simulation, noise-tuning alignment, and invariant representation learning inspired by the human visual cortex.
Next-generation AI Based on Neural Circuit Mechanism of Metacognition
Collaborated on large-scale neural-behavioral data collection and analysis using the AVATAR system.
Technical Guidance: SW Algorithm Test
Served as a technical advisor supervising the end-to-end design of software algorithm assessment problems.
ATM Cash Reserves Predictor
Developed a data-driven system to forecast ATM cash withdrawal demand, enabling optimal cash reserve allocation.