The More Popular, The Harder to Forget: Adaptive Popularity for LLM Unlearning

Benchmark Model Rank Results
sentence-completion-on-hellaswagAdaPop (Llama-3.1-8B-Instruct)#91Accuracy: 76.0