Our work Data poisoning for in-context learning is accepted to NAACL 2025!