List your company
ResearchFormerly WizWand

Memorization Reduction on GSM8K

35.2Memorization Reduction (%)GSS
-1.4088.09617.627.104Feb 9, 2026
Updated 8mo ago

Evaluation results

MethodLinks
GSSCategory=Ours, Model=Qwen, Utility threshold=>= 0.92026.0235.2
GSSCategory=Ours, Model=LLama, Utility threshold=>= 0.92026.0233.5
Task VectorCategory=Unlearning, Model=Qwen, Utility threshold=>= 0.92026.0225.4
Task VectorCategory=Unlearning, Model=LLama, Utility threshold=>= 0.92026.0225.1
ROMECategory=Editing, Model=Qwen, Utility threshold=>= 0.92026.0214.5
Balanced SubnetCategory=Editing, Model=Qwen, Utility threshold=>= 0.92026.0212.8
LUNARCategory=Editing, Model=LLama, Utility threshold=>= 0.92026.029
MEMITCategory=Editing, Model=Qwen, Utility threshold=>= 0.92026.028.7
ROMECategory=Editing, Model=LLama, Utility threshold=>= 0.92026.027.8
DPOCategory=Unlearning, Model=Qwen, Utility threshold=>= 0.92026.027.3
SimNPOCategory=Unlearning, Model=Qwen, Utility threshold=>= 0.92026.027.2
NPOCategory=Unlearning, Model=Qwen, Utility threshold=>= 0.92026.026.9
GACategory=Unlearning, Model=LLama, Utility threshold=>= 0.92026.024.5
SimNPOCategory=Unlearning, Model=LLama, Utility threshold=>= 0.92026.022.2
NPOCategory=Unlearning, Model=LLama, Utility threshold=>= 0.92026.022.2
MEMITCategory=Editing, Model=LLama, Utility threshold=>= 0.92026.022.2
Balanced SubnetCategory=Editing, Model=LLama, Utility threshold=>= 0.92026.021.4
DPOCategory=Unlearning, Model=LLama, Utility threshold=>= 0.92026.021.3
LUNARCategory=Editing, Model=Qwen, Utility threshold=>= 0.92026.021
GACategory=Unlearning, Model=Qwen, Utility threshold=>= 0.92026.020