Search papers, labs, and topics across Lattice.
Chao Chen and Chuanyi Liu are with Harbin Institute of Technology (Shenzhen). Yanhui Chen, Shanshan Lin, Dongsheng Hong, and Xiangwen Liao are with Fuzhou University. Shu Wu is with NLPR, MAIS, Institute of Automation, Chinese Academy of Sciences.Chao Chen and Chuanyi Liu were supported by the National Key Research and Development Program of China under Grant 2023YFB3106504. Chao Chen, Yanhui Chen, Shanshan Lin, Dongsheng Hong, and Xiangwen Liao were supported by National Natural Science Foundation of China under Grant 62476060. Shu Wu was supported by National Natural Science Foundation of China under Grant 62372454
2
0
4
LLMs can withstand 3,000 sequential knowledge edits without catastrophic forgetting, thanks to a new sparse editing framework that surgically manipulates knowledge circuits.
Get robust and interpretable models by combining adversarial training with explanation-guided learning, achieving a whopping +37% boost in adversarial accuracy on OOD data.
