Image Classification on ImageNet-C (val) Corruption Errors benchmark leaderboard · SOTA2 Research

Image Classification on ImageNet-C (val) Corruption Errors

27.4Error (Gaussian)CLIP

Evaluation Results

Method	Links
CLIP2026.04		27.4	—	39.6	—	—	28.7	34.6	25.3	41	36.7	49.8	44.1	49.7	65.4	35.1	30.3	53.5	42.2	29.4
CLIPArTT2026.04		28.2	—	40.2	—	—	29.8	35	26.5	41.4	37.3	50.3	43.7	49.7	65	38.5	31	53.5	42.3	30.2
Ensemble2026.04		29.1	—	41.6	—	—	30.1	37.5	27.3	44.2	39.2	52.4	46.4	52.6	67.8	34.4	32.4	56.2	44.2	30.4
DMN-ZS2026.04		29.2	—	41.7	—	—	30.2	37.6	27.6	44.4	39.2	52.5	46.6	52.7	67.9	33.8	32.5	56.5	44.5	30.5
TDA2026.04		29.6	—	42.7	—	—	31.5	38.1	28.9	44.6	40	53.4	47.5	53.3	68.5	39.1	33.4	57.1	45.1	31
WATT-S2026.04		31.8	—	43.7	—	—	33.6	39.2	30.7	45.7	41.3	53.5	47.3	53.7	67.9	40.1	34.7	57.3	45.6	33.1
Mint2026.04		33.8	—	44.5	—	—	35.8	39.3	33.3	46.2	41.5	53.6	47.4	53.3	67.7	36.3	38.2	58.1	47.4	35.2
Tent2026.04		36.3	—	46.7	—	—	38.5	40.3	37.4	47.7	43.5	54.2	49.5	56.4	67.8	45.4	40.6	57.6	46.6	38
NOTE2026.04		36.3	—	46.6	—	—	38.6	40	37.3	47.6	43.5	54.4	49.7	56.4	67.8	44.5	40.8	57.5	46.2	38
SAR2026.04		36.3	—	46.6	—	—	38.5	39.2	38.5	46.8	43.4	53.9	50.4	56.6	66.8	45.5	41.8	56.2	46.6	38.3
RoTTA2026.04		37.4	—	45.4	—	—	39.6	36.9	34.5	44	39.3	53.5	48.6	54.2	67.2	43	39.7	56.3	48.4	38.7
Ramen2026.04		37.5	—	49.2	—	—	41.1	42.2	39.6	49.9	46.2	57.3	51.6	59.4	68.7	46.4	45.2	59.4	54.7	38.8