Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Machine Unlearning research benchmarks – Page 17 · SOTA2 Research

Machine Unlearning

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Lacuna 1.0 (train test)

2.13Test Error

7

Feb 26, 2026

CIFAR-100 (Adversarial split (A))

86Epsilon Lower Bound

6

Jul 8, 2026

CIFAR-100 Uniform (U)

142.5Epsilon Lower Bound

6

Jul 8, 2026

6.45Time

6

Jun 24, 2026

6.33Time

6

Jun 24, 2026

87.92Accuracy (Forget)

6

Jun 1, 2026

Guardrail Classifier

0Refusal Rate

6

May 27, 2026

Guardrail Filter

2.6Refusal Rate

6

May 27, 2026

TOFU (Retain99)

Guardrail Filter

3.2Refusal Rate

6

May 27, 2026

LLaMA-3-8B Unlearning Evaluation Suite (test)

71.09Accuracy

6

May 12, 2026

94.61Retained Accuracy (Acc_r)

6

Apr 28, 2026

7Relearn Time

6

Apr 23, 2026

14Relearning Iterations

6

Apr 23, 2026

33Forget Rate

6

Apr 21, 2026

62Forget Rate

6

Apr 21, 2026

-1.1Forget Accuracy

6

Apr 14, 2026

Yelp (20 folds)

OLS θ_r^{(ols)}

0.12Running Time (s)

6

Apr 8, 2026

46.7ES

6

Apr 7, 2026

62.3ES

6

Apr 7, 2026

TOFU Llama3.1-8B (forget01)

99Forget Set RL Score

6

Mar 19, 2026

Complete Retraining

91.2Forgetting Rate

6

Mar 12, 2026

Complete Retraining

89.7Anxiety FR

6

Mar 12, 2026

CIFAR-10 (test)

90.8Plane Accuracy

6

Feb 26, 2026

CIFAR-100 10+10 classes (test)

74.3Retain Accuracy

6

Feb 26, 2026

60.5Forgetting Classes BF

6

Feb 27, 2026

Page 17 of 22

Popular datasets

CIFAR TOFU ImageNet WMDP MUSE

Follow for updates

Join SOTA2 Contact