Search papers, labs, and topics across Lattice.
Institute of Information Engineering, Chinese Academy of Sciences
1
0
1
Repurposing adversarial perturbations can significantly enhance continual learning, reducing forgetting and improving task transfer in large language models.
