LURK uses optimized adversarial suffixes and an LLM judge to show that unlearned models still leak Harry Potter-specific references at high rates.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Harry Potter is Still Here! Probing Knowledge Leakage in Targeted Unlearned Large Language Models via Automated Adversarial Prompting
LURK uses optimized adversarial suffixes and an LLM judge to show that unlearned models still leak Harry Potter-specific references at high rates.