Evaluation results

ForgetSet clip score of original model mean (~↑) on Forget set
self-reported

31.555
ForgetSet clip score of original model std (~↓) on Forget set
self-reported

1.602
ForgetSet clip score of unlearned model mean (↓) on Forget set
self-reported

31.574
ForgetSet clip score of unlearned model std (~↓) on Forget set
self-reported

2.340
ForgetSet clip score difference between original and unlearned mean (↑) on Forget set
self-reported

-0.019
ForgetSet clip score difference between original and unlearned std (~↓) on Forget set
self-reported

0.845
RetainSet clip score of original model mean (~↑) on Forget set
self-reported

32.258
RetainSet clip score of original model std (~↓) on Forget set
self-reported

3.672
RetainSet clip score of unlearned model mean (↑) on Forget set
self-reported

34.412
RetainSet clip score of unlearned model std (~↓) on Forget set
self-reported

1.910
RetainSet clip score difference between original and unlearned mean (↓) on Forget set
self-reported

-2.154
RetainSet clip score difference between original and unlearned std (~↓) on Forget set
self-reported

2.314
Inference latency seconds mean (↓) on Forget set
self-reported

1.883
Inference latency seconds std (~↓) on Forget set
self-reported

0.001
Runtime init seconds (~↓) on ['cat'] (forget) and ['lion', 'tiger', 'leopard'] (retain) sets
self-reported

0.000
Runtime data loading seconds (~↓) on ['cat'] (forget) and ['lion', 'tiger', 'leopard'] (retain) sets
self-reported

0.000
Runtime training seconds (↓) on ['cat'] (forget) and ['lion', 'tiger', 'leopard'] (retain) sets
self-reported

3.358
Runtime eval seconds (~↓) on ['cat'] (forget) and ['lion', 'tiger', 'leopard'] (retain) sets
self-reported

56.925