Most gradient-based LLM unlearning methods achieve behavioral suppression, not true forgetting, and current benchmarks cannot certify that knowledge has been removed.
Towards making systems forget with machine unlearning
1 Pith paper cite this work, alongside 20 external citations. Polarity classification is still indexing.
1
Pith paper citing it
20
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
LLM Unlearning for Cyber Defense: A Survey on Methods, Challenges, and Emerging Threats
Most gradient-based LLM unlearning methods achieve behavioral suppression, not true forgetting, and current benchmarks cannot certify that knowledge has been removed.