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

Image Classification on 7 Vision Benchmarks Suite (MNIST, SVHN, Cars, DTD, GTSRB, EuroSAT, RESISC45)

100MNIST Accuracy

Finetuning

Updated 22d ago

Evaluation Results

Method	Links
Finetuning 2026.07		100	100	100	100	100	100	100	100
Finetuning 2026.07		100	100	100	100	100	100	100	100
PoE-EBM 2026.07		99.68	96.45	94.91	85.47	95.18	97.14	92.21	94.43
Concrete-TA 2026.07		99.23	90.38	91.82	68.06	87.9	94.03	92.38	89.12
PoE-EBM 2026.07		99.2	94.03	86.34	78.19	89.6	90.24	81.64	88.46
DOGE-TA 2026.07		99.12	96.49	97.49	73.88	94.8	96.55	95.01	93.34
TA-DARE-TIES 2026.07		98.86	90.04	89.48	74.37	84.43	87.91	86.3	87.36
DOGE-TA 2026.07		98.73	89.8	92.37	66.1	88.55	90.13	86.3	87.42
Concrete-TA 2026.07		97.31	82.09	75.95	60.2	71.92	76.39	75.01	77.08
TA-DARE-TIES 2026.07		96.29	84.16	77.81	62.6	71.48	70.64	70.75	76.25
Task Arithmetic (TA) 2026.07		93.12	72.18	86.34	61.64	66.96	69.03	76.82	75.16
Task Arithmetic (TA) 2026.07		90.95	73.52	78.81	56.53	63.25	66.3	72.53	71.7
Fisher Merging 2026.07		90.06	84.49	90.56	56.04	58.5	77.28	77.57	75.91
Weight averaging 2026.07		85.28	65.71	78.71	53.2	55.89	63.59	71.56	67.7
Fisher Merging 2026.07		83.94	63.58	86.24	58.65	59.03	99.18	71.51	74.89
Weight averaging 2026.07		80.05	60.42	83.91	56.15	57.83	54.6	70.42	66.2