I'm a third-year PhD student in Computer Science at Georgia Tech, advised by Christopher MacLellan. I'm also a Research Intern at Amazon in Washington, working on deployment-time agent adaptation and self-improving harnesses.
I study continual learning for deployment-time adaptation and generalization.
Selected papers
Zekun Wang, Anant Gupta, Christopher J. MacLellan
NeurIPS 2026
Combines generative replay with a Fisher-metric trust region for continual diffusion training, and shows the update reduces to a one-step MAML form.
Zekun Wang, Yingqiang Ge, Hai Wang, Yuhui Xu, Joshua Frandsen, Eric Yang, Shancong Fu, Christopher J. MacLellan, Chandan K. Reddy
Preprint 2026
A proposer meta-trained across agent benchmarks reads a few failures of an unseen benchmark and writes one harness edit that improves the agent there.
Zekun Wang, Anant Gupta, Zihan Dong, Christopher J. MacLellan
arXiv 2026
SCoL post-trains an LLM via meta-RL to emit textual instructions selecting which Transformer layers to LoRA-update from streaming context, balancing acquisition against forgetting.
Zekun Wang, Anant Gupta, Tianyi Zhu, Christopher J. MacLellan
arXiv 2026
Recovers prototypes from a pretrained diffusion model by score ascent at several noise levels, and composes those that explain a single out-of-distribution query into a product of experts at test time.
Zekun Wang, Anant Gupta, Zihan Dong, Christopher J. MacLellan
ICLR 2026
Proposes a rank-1 Fisher approximation for EWC that exploits the collinear per-sample gradients of diffusion models at low SNR, paired with generative replay for continual image generation.
Ziqiao Ma, Zekun Wang, Joyce Chai
NAACL 2025
Trains GPT-2 from scratch with a teacher model giving demonstrations on student trials, scored by an age-conditioned reward, to accelerate word acquisition.
All papers