LLM Unlearning
Benchmarks
DatasetSOTA methodMetricTrendResultsLast updated
100Accuracy
30
Apr 16, 2026
-13.591FQ
25
Feb 26, 2026
44Forget Rate
24
Apr 21, 2026
82.5DF ROUGE
18
Jun 1, 2026
5.62Fluency
18
Feb 26, 2026
57.19In-Domain Accuracy
18
Feb 26, 2026
61.96Accuracy
18
Feb 26, 2026
22.09Verification Memory (VerMem)
16
Feb 26, 2026
0VerMem
16
Feb 26, 2026
89.3USR
16
Feb 26, 2026
62.4Delta Score
14
May 8, 2026
1FQ
13
Jun 16, 2026
93.2DF ROUGE
13
Jun 1, 2026
61Aggregated Score
13
May 8, 2026
0.686Model Utility
11
Jun 11, 2026
90.8DF ROUGE
10
Jun 1, 2026
68.5Forget Rate
8
Apr 20, 2026
89USR
8
Feb 26, 2026
91.8USR
8
Feb 26, 2026
91.6USR
8
Feb 26, 2026
6.57FQ
7
Jun 11, 2026
55.8R-MCQA
6
Jun 11, 2026
5.04F-Cloze
6
Jun 11, 2026
7.456RQ
5
May 19, 2026
7.456RQ
5
May 19, 2026