PhD Candidate in Machine Learning at UNSW Sydney, supervised by Prof. Dong Gong.
My research studies how models can learn, remember, and reason reliably under constraints — when data is scarce, when past information can't be stored, and when adaptation must happen without updating model weights. I work across continual learning, data-efficient learning, and self-improving LLM reasoning.
🔭 Research interests: continual learning · coreset selection · model inversion · LLM reasoning · memory-augmented and self-improving agents
- MILES — Modular Instruction Memory with Learnable Selection for Self-Improving LLM Reasoning · project page
- PMI-CFS — Model Inversion with Layer-Specific Modeling and Alignment for Data-Free Continual Learning · NeurIPS 2025 · code
- CSReL — Coreset Selection via Reducible Loss in Continual Learning · ICLR 2025 · code