I am an M.Phil. researcher in Artificial Intelligence at HKUST(GZ). I study how language models represent, transport, and revise reasoning, and I build causal tests and reliable interventions for latent reasoning, retrieval, and model evaluation.

• Causal mechanisms of latent reasoning — SCIT, causal cache carriers, and transport versus destructive sensitivity
• Reliable retrieval and representation intervention — FCPRAG, target selection margins, and causal SAE retrieval
• Efficient reasoning systems — inference-time control, collaborative reasoning, and rigorous evaluation

• SCIT — first-author paper accepted at EMNLP 2026
• FCPRAG — accepted at EMNLP 2026
• MARIO MICCAI 2024 — sole-author work, Best Method and third place

• Research website
• English and Chinese CV
• Email