TOFU
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
TOFU Forget Set
68.28FB
17
TOFU forget01 LLaDA-8B-Base
35.55ForgetLP
16
TOFU 20% known forget set (DFuk)
17.35Recovery Rate
15
TOFU 10%
94.8ROUGE-L Score (Forget)
14
TOFU
100FT Score
14
TOFU finetuned Llama-2-7b-chat (forget set)
99.51Probability
14
TOFU Llama3.1-8B (forget10)
98Forgetting Quality (FQ)
14
TOFU (Forget)
0.986ROUGE
14
TOFU Real World
89.7ROUGE
14
TOFU (Real Authors)
93.3ROUGE
14
TOFU (forget01)
1FQ
13
TOFU
93.2DF ROUGE
13
TOFU
61Aggregated Score
13
TOFU forget10
35Time (s)
13
TOFU Forget10 Llama-3.2-1B
100MEM Score
12
TOFU Llama 2 1.0 (forget 10%)
85.14UFE
12
TOFU Llama 2 1.0 (forget 5%)
88.11UFE Score
12
TOFU Llama 2 1.0 (forget 1%)
92.02UFE
12
TOFU Phi 1.5 1.0 (forget 10%)
66.88UFE
12
TOFU Phi 1.5 1.0 (forget 1%)
74.62UFE
12
TOFU
83Forget Rate
12
TOFU Llama 3.1 8B (5%)
0.914FQ
12
TOFU
61.4Model Utility
12
TOFU 1% forget
54Cumulative ROUGE-1
12
ToFU General Knowledge (GEK) (one author (20 samples))
90Accuracy
11