Multi-Task Incremental Learning on MTIL few-shot Order II (test) benchmark leaderboard · SOTA2 Research

Multi-Task Incremental Learning on MTIL few-shot Order II (test)

88.8Accuracy (Food)MoE-Adapters

Evaluation Results

Method	Links
MoE-Adapters2026.05		88.8	64.7	—	—	—	59.5	89.1	71.2	65.3	18.2	87.9	44.2	54.6	68.2	64.7
Zero-shot2026.05		88.5	—	—	—	64.7	59.4	89	71	65.2	24.3	88.4	44.6	54.9	68.2	65.3
DIMoE-Adapters2026.05		88.5	65.4	—	—	—	58.9	89.1	71	65.9	24.4	88.5	44.7	55	68.2	65.4
DIMoE-Adapters2026.05		87.8	—	71.4	—	70.6	85.2	90.4	83.2	69.7	29.6	89.4	49.7	60.8	69	71.4
DIMoE-Adapters2026.05		87.7	—	—	77.9	70.5	91.1	90.9	90.4	72.9	35.8	91	62.6	86.8	77	77.9
ZSCL2026.05		87.3	62.8	—	—	—	64.8	85.3	67.9	64.5	18.9	86.6	43.6	43.2	65.7	62.8
MoE-Adapters2026.05		87	—	69.5	—	61.2	87.3	89.1	79.3	68.5	23.4	89.4	49.9	60.8	68.8	69.5
MoE-Adapters2026.05		87	—	—	75.7	59.4	91.8	89	84.1	71.9	29.4	91.4	64.2	88.8	75	75.7
ZSCL2026.05		85	—	67.6	—	70	79.8	86.1	79.4	68.3	21.8	88.8	48.8	49.3	66.5	67.6
5-shot Full Fine-tune2026.05		83.3	—	—	—	65.4	96.6	84.9	92.9	71.3	30.6	93.5	65.1	91.7	76.8	77.5
ZSCL2026.05		82.7	—	—	71.8	67.4	78.7	85.7	81.3	71.2	25	92.5	62	72.2	74.4	71.8
LwF-VR2026.05		80.1	54.4	—	—	—	55.4	77.7	50.4	61.4	9.1	83.5	40.1	31.5	54.8	54.4
WiSE-FT2026.05		77.3	53.7	—	—	—	60	76.9	54.2	58	11.1	81.8	37.6	31.7	48.1	53.7
LwF-VR2026.05		76.9	—	60.1	—	63.3	71.4	79.1	68.9	65	13.4	86	45.7	36.3	55.3	60.1
LwF-VR2026.05		70.1	—	—	63.5	57.3	72.1	74.6	71.9	65.8	17.4	89.5	60	56	60.2	63.5
WiSE-FT2026.05		64.7	—	55.1	—	59.3	77.4	70.3	51.3	58.6	10.8	84.2	42	38.6	49.1	55.1
LwF2026.05		64.2	48.6	—	—	—	59.1	68.1	38.4	54.9	6.7	78	35.5	33.5	47.4	48.6
LwF2026.05		55	—	54.7	—	64.1	79.5	69.2	55.7	58.3	10.8	81.7	41.3	39.2	47.4	54.7
WiSE-FT2026.05		47.7	—	—	51.8	48.1	66.9	59.8	25	56.1	7.4	88.5	52.2	66.8	59.4	51.8
LwF2026.05		40.1	—	—	56.8	57.1	84.1	58.1	50.5	57.6	14.3	87.9	54.7	64	47	56.8