Concept Unlearning on Prompt Attacks benchmark leaderboard · SOTA2 Research

Concept Unlearning on Prompt Attacks

35.7ASR (ORIG)ISPF

Evaluation Results

Method	Links
ISPF2026.06		35.7	28.6	8.6	45.2	28.5	0	24
GA2026.06		8.5	3	4	7	16	0	6.4
GA+KL2026.06		8	3	3.5	7	16.5	0	6.3
SIU2026.06		4.5	2.8	2.3	4.5	10	0	4
SPACE2026.06		4	4	4.5	8	10	0	5.1
NPO2026.06		1	2.5	1	2	3.5	0	1.7