1 paper
Yanda Chen, Chen Zhao, Zhou Yu +2
In-context learning (ICL) suffers from oversensitivity to the prompt, making it unreliable in real-world scenarios. We study the sensitivity of ICL with respect to multiple perturb…