Share your thoughts, 1 month free Claude Pro on usSee more

Image Classification on VTAB

91.97Overall Accuracy

Full-Tune

Updated 1mo ago

Evaluation Results

Method
Full-Tune 2026.04	91.97	-	-	-	-	-
Full-Tune 2026.04	91.97	-	-	-	-	-
AdapterFormer 2026.04	90.69	-	-	-	-	-
AdapterFormer 2026.04	90.69	-	-	-	-	-
AP 2026.04	90.25	-	-	-	-	-
AP 2026.04	90.25	-	-	-	-	-
E2VPT 2026.04	90.12	-	-	-	-	-
VPT 2026.04	89.97	-	-	-	-	-
Bias 2026.04	89.84	-	-	-	-	-
Bias 2026.04	89.84	-	-	-	-	-
LoRA 2026.04	89.81	-	-	-	-	-
LoRA 2026.04	89.81	-	-	-	-	-
Adapter 2026.04	89.44	-	-	-	-	-
Adapter 2026.04	89.44	-	-	-	-	-
SSF 2026.04	89.43	-	-	-	-	-
SSF 2026.04	89.43	-	-	-	-	-
VPT 2026.04	89.24	-	-	-	-	-
GateVPT 2026.04	89.14	-	-	-	-	-
Linear-Probe 2026.04	78.9	-	-	-	-	-
Linear-Probe 2026.04	78.9	-	-	-	-	-
Franca 2026.03	72.1	78.6	84.5	60.1	-	-
DINOv2 (dst) 2026.03	71.2	79.6	84.7	57	-	-
DINOv2 (dst) 2026.03	71.1	78.1	84.2	58.3	-	-
CAPI 2026.03	70.9	72.7	83.5	63.1	-	-
Franca 2026.03	69.3	75.1	83.3	57.3	-	-
DINOv2 (dst) 2026.03	69.1	74.1	83	57.7	-	-
Bootleg 2026.03	68.7	70.9	84.2	58.9	-	-
MAE 2026.03	67.5	65.7	81.6	62.1	-	-
CrossMAE 2026.03	67.3	65.8	81	61.8	-	-
Bootleg 2026.03	67	68	83.7	57.8	-	-
MAE 2026.03	65.1	64.4	79.2	58.7	-	-
CrossMAE 2026.03	65.1	68.4	81.4	54.1	-	-
data2vec 2.0 2026.03	63.6	55.5	80.1	62.5	-	-
Bootleg 2026.03	63.5	66.2	80.8	52.5	-	-
CrossMAE 2026.03	62.8	59.7	78.8	57.5	-	-
DINOv2 (dst) 2026.03	62.4	72.9	83.8	42.5	-	-
DINOv2 (dst) 2026.03	61.8	72.8	83.1	41.3	-	-
DINOv2 (dst) 2026.03	61.5	75.5	84.4	37.7	-	-
DINOv2 (dst) 2026.03	60.3	74.6	81.7	37.2	-	-
MAE 2026.03	60.1	57.6	78.2	53.2	-	-
Franca 2026.03	59.9	71.2	82.2	38.9	-	-
DINOv2 (dst) 2026.03	59.1	69	81.6	39.3	-	-
DINOv2 (dst) 2026.03	59	72.1	81.1	36.4	-	-
CAPI 2026.03	58.9	62.4	80.8	44.9	-	-
UpperCaSE 2022.06	58.4	69.1	80.3	39.4	-	-
Franca 2026.03	58.1	66.5	82.1	38.6	-	-
data2vec 2.0 2026.03	58	54.6	72.4	53.8	-	-
MAE 2026.03	57.9	60.1	80.2	44.7	-	-
I-JEPA 2026.03	57.8	55.1	78.5	49.8	-	-
CrossMAE 2026.03	57.8	57.6	80.5	46.5	-	-
CrossMAE 2026.03	57.7	59.1	78.8	46	-	-
I-JEPA 2026.03	57.5	58	79.1	46.3	-	-
Bootleg 2026.03	57.4	64.9	79.7	39.7	-	-
Bootleg 2026.03	57.1	60.7	81.9	41.5	-	-
Franca 2026.03	56.9	71.3	79.7	32.8	-	-
UpperCaSE 2022.06	56.6	66.3	80.1	37.6	-	-
MAE 2026.03	56.6	56.2	79.4	45.4	-	-
data2vec 2.0 2026.03	56.5	48.4	74.1	54.7	-	-
CrossMAE 2026.03	56.4	56.4	77.6	45.7	-	-
Bootleg 2026.03	56.4	60.7	82.1	39.8	-	-
Bootleg 2026.03	55.6	59.8	79	40.4	-	-
Bootleg 2026.03	55.6	61.5	80.5	38	-	-
Franca 2026.03	55.5	68.7	78.9	32.2	-	-
data2vec 2.0 2026.03	55	51.9	77.8	46.3	-	-
Bootleg 2026.03	54.9	59.8	78.1	38.9	-	-
CrossMAE 2026.03	54.8	55.2	74.3	44.7	-	-
MAE 2026.03	53.9	54.3	76.7	42.2	-	-
data2vec 2.0 2026.03	53.9	52.6	76.6	43.7	-	-
MAE 2026.03	53.6	53.4	76.8	42.2	-	-
data2vec 2.0 2026.03	52.8	47.9	74	46.5	-	-
I-JEPA 2026.03	52.4	54	78.4	38	-	-
I-JEPA 2026.03	51.6	50.4	77.2	39.8	-	-
LITE 2022.06	51.4	65.2	71.9	30.8	-	-
CrossMAE 2026.03	51.3	47.9	77.4	41.2	-	-
I-JEPA 2026.03	50.9	45.8	73.8	43.9	-	-
CTX 2022.06	50.5	61.1	67.3	34	-	-
data2vec 2.0 2026.03	50.2	43.9	71.6	44.9	-	-
CrossMAE 2026.03	50.1	46.4	76	40.2	-	-
MAE 2026.03	49.7	46.2	77.3	38.9	-	-
MAE 2026.03	49.5	45.7	76.6	39.1	-	-
data2vec 2.0 2026.03	46.9	38.8	74.6	40.1	-	-
I-JEPA 2026.03	46.3	40.2	74.3	37.7	-	-
ProtoNet 2022.06	46.1	60.9	64.2	25.9	-	-
data2vec 2.0 2026.03	45.8	37.4	75.6	38.4	-	-
ProtoMAML 2022.06	45	45.7	70.7	31.5	-	-
DFN ∪ Ours (20%)* 2024.05	38.6	-	-	-	-	-
DFN ∪ HYPE 2024.05	38.5	-	-	-	-	-
DFN ∪ HYPE ∪ Ours (10%)* 2024.05	38.5	-	-	-	-	-
DFN 2024.05	36.2	-	-	-	-	-
MLM 2024.05	36	-	-	-	-	-
Devil 2024.05	35.9	-	-	-	-	-
Ours (20%) 2024.05	35.9	-	-	-	-	-
T-MARS 2024.05	34.8	-	-	-	-	-
HYPE 2024.05	34.3	-	-	-	-	-
TripletCLIP 2024.11	20.81	-	-	-	53.4	-
Self-Filtering 40% 2026.06	19.7	-	-	-	-	-
CLIP with OAI CLIP Filtered data + All 2026.06	19.6	-	-	-	-	-
NegCLIP 2024.11	19.12	-	-	-	50.56	-
Final Self-Filtered 30% Data Mix 2026.06	19.1	-	-	-	-	-
LaCLIP 2024.11	19.08	-	-	-	49.06	-

Showing 100 of 125 rows