Maochuan Lu — accepted AI-conference papers

2025

Efficient Knowledge Editing via Minimal Precomputation

Akshat Gupta, Maochuan Lu, Thomas Hartvigsen, Gopala Anumanchipalli

ACL 2025short

Knowledge editing methods like MEMIT are able to make data and compute efficient updates of factual knowledge by using a single sentence to update facts and their consequences. However, what is often overlooked is a “precomputation step”, which requires a one-time but significant computational cost.…

Cited by 0 Source ↗PDF ↗Scholar ↗

2025

Lifelong Knowledge Editing requires Better Regularization

Akshat Gupta, Phudish Prateepamornkul, Maochuan Lu, Ahmed Alaa, Thomas Hartvigsen, Gopala Anumanchipalli

EMNLP 2025

Knowledge editing is a promising way to improve factuality in large language models, but recent studies have shown significant model degradation during sequential editing. In this paper, we formalize the popular locate-then-edit methods as a two-step fine-tuning process, allowing us to precisely ide

Cited by 0 Source ↗PDF ↗code ↗Scholar ↗