Knowledge Unlearning on 16-task Sequential Unlearning Forgotten Data Avg benchmark leaderboard · SOTA2 Research

Knowledge Unlearning on 16-task Sequential Unlearning Forgotten Data Avg

90.9CRRCORE

Evaluation Results

Method	Links
CORE2026.03		90.9	3.6	43.42	53.83	6.08	—	—	—	—
CORE2026.03		85.22	7.08	—	—	—	96.32	84.27	85.57	44.86
O32026.03		82.55	6.44	45.91	57.92	9.32	—	—	—	—
O32026.03		75	12.07	—	—	—	93.04	75.72	81.24	48.04
MoEAdapter2026.03		66.27	22.96	43.93	55.2	4.83	—	—	—	—
EWC2026.03		55.21	42.21	—	—	—	77.28	27.83	47.22	42.41
GMM2026.03		55.13	44.22	—	—	—	65.65	18.9	44.58	46.3
SCRUB2026.03		54.53	44.76	—	—	—	66.61	19.81	44.85	46.29
EProj2026.03		53.94	43.37	—	—	—	73.28	26.88	47.36	42.69
LwF2026.03		53.88	43.23	—	—	—	83.71	39.03	51.14	48.59
MoEAdapter2026.03		45.2	54.26	—	—	—	62.17	28.52	51.98	36.28
EProj2026.03		25	45.83	50.28	59.21	13.26	—	—	—	—
SCRUB2026.03		23.3	53.02	47.81	57.2	11.84	—	—	—	—
GMM2026.03		22.3	52.64	48.08	57.56	11.99	—	—	—	—
EWC2026.03		18.92	45.86	52.15	61.16	15.04	—	—	—	—
LwF2026.03		13.94	47.57	51.32	60.45	14.53	—	—	—	—
Llama-22026.03		11.34	12.88	—	—	—	100	82.8	100	100
Vicuna2026.03		1.91	2.01	100	100	100	—	—	—	—