Machine Unlearning
Benchmarks
DatasetSOTA methodMetricTrendResultsLast updated
99.94Accr
46
Jun 4, 2026
93ROUGE-L
45
May 19, 2026
99.9Forgetting Accuracy (Train)
43
Mar 30, 2026
1Forget Quality (FQ)
43
Feb 26, 2026
0.99Forget Quality - Rouge-L
42
Apr 20, 2026
0.99Forget Quality (ROUGE-L)
42
Mar 19, 2026
99.98RA (Utility Retention)
40
Apr 14, 2026
0KL Divergence (t)
39
May 12, 2026
85.9FB Score
39
Feb 26, 2026
100Retain Accuracy
37
May 13, 2026
18.511Avg Increase in Forget Accuracy
36
Mar 10, 2026
95.41Test Acc
35
May 19, 2026
58.42VerbMem (Df)
34
Jun 25, 2026
0.884ROUGE-L
34
May 19, 2026
89Forget Quality
33
Jun 4, 2026
92Utility Preservation
33
May 15, 2026
51.52Model Utility (MU)
32
May 22, 2026
78.975Time-over-Weight (High)
32
Mar 10, 2026
14.32Log Forget Quality (-log(FQ))
30
Apr 21, 2026
2.896-log(FQ)
30
Apr 21, 2026
13.955-log(FQ)
30
Apr 21, 2026
99.702No VerbMem
30
Apr 29, 2026
-396.8fLP
30
Jun 12, 2026
61.15MMLU
29
Apr 28, 2026
53.7BF16 Score
28
May 15, 2026