Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Machine Unlearning on Harry Potter (test)
Loading...
74.8
BLEU
Original
-2.992
17.204
37.4
57.596
Mar 2, 2026
BLEU
ROUGE-L
ASG
Updated 1mo ago
Evaluation Results
Method
Method
Links
BLEU
ROUGE-L
ASG
Original
Base Model=Llama-2-7B
2026.03
74.8
85.1
74.1
WHP
Base Model=Llama-2-7B
2026.03
23.6
17.9
14.9
ELM
Base Model=Llama-2-7B
2026.03
8.1
9
2.7
LoRA
Base Model=Llama-2-7B
2026.03
7.2
11.5
3.5
Fine-tune
Base Model=Llama-2-7B
2026.03
6.4
17.2
5.9
A-LoRA
Base Model=Llama-2-7B
2026.03
5.9
10.4
1.9
Ours
Base Model=Llama-2-7B
2026.03
4.7
9.6
1.3
GD
Base Model=Llama-2-7B
2026.03
3.9
14.5
3.4
Retain
Base Model=Llama-2-7B
2026.03
1.9
9.8
0
NPO
Base Model=Llama-2-7B
2026.03
1.5
5.3
2.5
KL
Base Model=Llama-2-7B
2026.03
1.2
8.9
0.8
GA
Base Model=Llama-2-7B
2026.03
0
0
6
Feedback
Search any
task
Search any
task
