List your company

Image Classification on Tiny-ImageNet (Accuracy %)

86.93Accuracy (%)LoRSP
24.186840.475956.76573.0541Jan 12, 2025Apr 7, 2025Jul 1, 2025Sep 24, 2025Dec 18, 2025Mar 13, 2026Jun 7, 2026

Evaluation results

MethodLinks
LoRSPPre-trained Backbone=ViT-B/322026.0686.93
HIVEFoundation Model=CLIP2026.0386.71
AS-LoRADP Budget (ϵ)=32026.0586.66
BaseFoundation Model=CLIP2026.0386.65
SAFoundation Model=CLIP2026.0386.51
LoR-VPPre-trained Backbone=ViT-B/322026.0685.77
FFA-LoRADP Budget (ϵ)=32026.0585.5
RoLoRADP Budget (ϵ)=32026.0584.62
LPPre-trained Backbone=ViT-B/322026.0683.95
AS-LoRADP Budget (ϵ)=12026.0583.71
VPTPre-trained Backbone=ViT-B/322026.0683.54
FFA-LoRADP Budget (ϵ)=12026.0583.52
HIVEFoundation Model=SigLIP2026.0382.48
SAFoundation Model=SigLIP2026.0382.46
AutoVPPre-trained Backbone=ViT-B/322026.0682.43
RoLoRADP Budget (ϵ)=12026.0582.34
BaseFoundation Model=SigLIP2026.0382.27
LSFormerArchitecture=LSFormer-4-384, Timesteps=42026.0571.61
Spikinghash-MArchitecture=Spikinghash-M, Param (M)=6.28, Time Step=42025.0168.26
SAFormerArchitecture=SAFormer-4-384, Timesteps=42026.0567.28
Spikingformer-CMLArchitecture=Spikingformer, Param (M)=8.40, Time Step=42025.0166.59
SpikformerArchitecture=Spikformer-4-384, Timesteps=42026.0566.03
S-TransformerArchitecture=Spikingformer-4-384, Timesteps=42026.0565.71
SpikformerArchitecture=Spikformer, Param (M)=9.40, Time Step=42025.0165.7
LM-HT(L=4)Architecture=VGG-13, Timesteps=22026.0562.05
BPLearning Paradigm=BP-based2026.0661.8
LM-HT(L=2)Architecture=VGG-13, Timesteps=42026.0561.75
LM-HArchitecture=VGG-13, Timesteps=42026.0559.51
DCTArchitecture=VGG-13, Timesteps=1252026.0556.9
ASGLArchitecture=VGG-13, Param (M)=60.56, Time Step=82025.0156.81
ASGLArchitecture=VGG-13, Timesteps=42026.0556.57
TeacherArchitecture=ResNet322026.0456.27
Offline-LTArchitecture=VGG-13, Param (M)=60.56, Time Step=162025.0155.37
Offline-LTLArchitecture=VGG-13, Timesteps=162026.0555.37
Online-LTLArchitecture=VGG-13, Param (M)=60.56, Time Step=162025.0154.82
Online-LTLArchitecture=VGG-13, Timesteps=162026.0554.82
QCFSArchitecture=VGG-16, Param (M)=65.87, Time Step=322025.0153.54
Student-FullArchitecture=ResNet202026.0451.51
Proposed methodLearning Paradigm=Forward-only2026.0649.87
PODSSelection Ratio=70%, Backbone=ResNet-502026.0549
Dyn. Sel. Aug.Selection Ratio=70%, Backbone=ResNet-502026.0548.5
PODSSelection Ratio=50%, Backbone=ResNet-502026.0546.9
CLIP-SelSelection Ratio=70%, Backbone=ResNet-502026.0546
FAPDalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0345.35
FAPDncontalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0345.28
Dyn. Sel. Aug.Selection Ratio=50%, Backbone=ResNet-502026.0545.1
Full DatasetSelection Ratio=100%, Backbone=ResNet-502026.0545
FedRCLalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0344.89
FAPDnadptalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0344.67
FedCDAalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0344.56
RL-SelectorSelection Ratio=70%, Backbone=ResNet-502026.0544.5
PODSSelection Ratio=30%, Backbone=ResNet-502026.0544.5
DivBFKDDistillation setting=black-box, Real image budget (N)=10 K, Synthetic image budget (M)=50 K2026.0444.25
FedNHalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0344.12
EL2NSelection Ratio=70%, Backbone=ResNet-502026.0544
GlisterSelection Ratio=70%, Backbone=ResNet-502026.0543.9
InfoBatchSelection Ratio=70%, Backbone=ResNet-502026.0543.8
MOONalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0343.78
FedProxalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0343.48
Self-sup. proto.Selection Ratio=70%, Backbone=ResNet-502026.0543.4
MoSoSelection Ratio=70%, Backbone=ResNet-502026.0543.4
FedAvgalpha=0.5, Backbone=MobileNetV2, Local Epochs (E)=10, Number of Clients=102026.0343.35
FS-BBTDistillation setting=black-box, Real image budget (N)=10 K, Synthetic image budget (M)=50 K2026.0443.29
GraNdSelection Ratio=70%, Backbone=ResNet-502026.0543.2
InfoBatchSelection Ratio=50%, Backbone=ResNet-502026.0543.2
Random*Selection Ratio=70%, Backbone=ResNet-502026.0543.1
Moderate-DSSelection Ratio=70%, Backbone=ResNet-502026.0542.8
Random*Selection Ratio=50%, Backbone=ResNet-502026.0542.8
DiT-OursIPC=502026.0642.8
RandomSelection Ratio=70%, Backbone=ResNet-502026.0542.2
InfoBatchSelection Ratio=30%, Backbone=ResNet-502026.0542.2
Random*Selection Ratio=30%, Backbone=ResNet-502026.0541.5
Dyn. Sel. Aug.Selection Ratio=30%, Backbone=ResNet-502026.0541.5
DeeperForwardLearning Paradigm=Forward-only, Reproduced results=true2026.0641.36
ForgettingSelection Ratio=70%, Backbone=ResNet-502026.0541.2
Standard-KDDistillation setting=black-box, Real image budget (N)=10 K, Synthetic image budget (M)=50 K2026.0440.42
BBKDDistillation setting=black-box, Real image budget (N)=10 K, Synthetic image budget (M)=50 K2026.0440.01
CLIP-SelSelection Ratio=50%, Backbone=ResNet-502026.0540
HerdingSelection Ratio=70%, Backbone=ResNet-502026.0539.8
GlisterSelection Ratio=50%, Backbone=ResNet-502026.0539.5
MoSoSelection Ratio=50%, Backbone=ResNet-502026.0538.5
Moderate-DSSelection Ratio=50%, Backbone=ResNet-502026.0538.2
Self-sup. proto.Selection Ratio=50%, Backbone=ResNet-502026.0537.9
RandomSelection Ratio=50%, Backbone=ResNet-502026.0537.2
EL2NSelection Ratio=50%, Backbone=ResNet-502026.0537.1
GraNdSelection Ratio=50%, Backbone=ResNet-502026.0536.3
DiT-OursIPC=102026.0635.9
ForgettingSelection Ratio=50%, Backbone=ResNet-502026.0533
CLIP-SelSelection Ratio=30%, Backbone=ResNet-502026.0531.7
HerdingSelection Ratio=50%, Backbone=ResNet-502026.0531.6
MoSoSelection Ratio=30%, Backbone=ResNet-502026.0531.2
RL-SelectorSelection Ratio=30%, Backbone=ResNet-502026.0531.1
Moderate-DSSelection Ratio=30%, Backbone=ResNet-502026.0530.6
GlisterSelection Ratio=30%, Backbone=ResNet-502026.0530.1
RandomSelection Ratio=30%, Backbone=ResNet-502026.0529.8
GraNdSelection Ratio=30%, Backbone=ResNet-502026.0529.7
HerdingSelection Ratio=30%, Backbone=ResNet-502026.0529.4
ForgettingSelection Ratio=30%, Backbone=ResNet-502026.0528.7
Self-sup. proto.Selection Ratio=30%, Backbone=ResNet-502026.0527.7
EL2NSelection Ratio=30%, Backbone=ResNet-502026.0526.6