Image Classification on CIFAR-10 (Accuracy, Time Saved, Annotation Simulation Time)
96.18AccuracyEnsemble AL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Ensemble ALModel=ResNet-18, Candidate Model Used?=Yes2026.05 | 96.18 | 0 | — | |
| Low Confidence (LC)Model=VGG-16, Candidate Model Used?=No2026.05 | 94.21 | 0.5 | 8.77 | |
| High Confidence and Least Confidence (HCLC)Model=VGG-16, Candidate Model Used?=No2026.05 | 94.2 | 0.5 | 8.77 | |
| Low Confidence (LC)Model=MobileNetV2, Candidate Model Used?=No2026.05 | 94.16 | 0.2 | 6.52 | |
| High Confidence and Least Confidence (HCLC)Model=MobileNetV2, Candidate Model Used?=No2026.05 | 94.16 | 0.2 | 6.52 | |
| High Confidence (HC)Model=VGG-16, Candidate Model Used?=No2026.05 | 93.94 | 0.5 | 8.77 | |
| PruneFuseModel=ResNet-56, Candidate Model Used?=Yes2026.05 | 93.65 | 0 | — | |
| Low Confidence (LC)Model=ResNet-18, Candidate Model Used?=No2026.05 | 93.53 | 0.25 | 4.27 | |
| High Confidence and Least Confidence (HCLC)Model=ResNet-18, Candidate Model Used?=No2026.05 | 93.48 | 0.25 | 4.27 | |
| High Confidence (HC)Model=ResNet-18, Candidate Model Used?=No2026.05 | 93.28 | 0.25 | 4.27 | |
| High Confidence and Least Confidence (HCLC)Model=DenseNet-121, Candidate Model Used?=No2026.05 | 93.08 | 0.16 | 5.09 | |
| Low Confidence (LC)Model=DenseNet-121, Candidate Model Used?=No2026.05 | 92.87 | 0.16 | 5.09 | |
| GlisterModel=DenseNet-121, Candidate Model Used?=Yes2026.05 | 92.86 | 0 | 9.49 | |
| High Confidence (HC)Model=MobileNetV2, Candidate Model Used?=No2026.05 | 92.84 | 0.2 | 6.52 | |
| High Confidence (HC)Model=DenseNet-121, Candidate Model Used?=No2026.05 | 92.72 | 0.16 | 5.09 | |
| SSALModel=ResNet-18, Candidate Model Used?=Yes2026.05 | 92.7 | 0 | — | |
| Low Confidence (LC)Model=ResNet-18, Candidate Model Used?=No, Active Learning Budget=40K2026.05 | 92.69 | 0.25 | 3.4 | |
| High Confidence (HC)Model=ResNet-18, Candidate Model Used?=No, Active Learning Budget=40K2026.05 | 92.69 | 0 | 3.4 | |
| Band-limited TrainingModel=DenseNet-121, Candidate Model Used?=No2026.05 | 92 | — | — | |
| PrACModel=ResNet-56, Candidate Model Used?=Yes2026.05 | 92 | 0 | — | |
| Low Confidence (LC)Model=VGG-16, Candidate Model Used?=No, Active Learning Budget=40K2026.05 | 91.89 | 0.5 | 6.25 | |
| Yun et al.Model=DenseNet-121, Candidate Model Used?=Yes2026.05 | 91.7 | 0 | — | |
| Low Confidence (LC)Model=ResNet-56, Candidate Model Used?=No2026.05 | 91.6 | 0.26 | 8.33 | |
| High Confidence and Least Confidence (HCLC)Model=ResNet-56, Candidate Model Used?=No2026.05 | 91.6 | 0.26 | 8.33 | |
| CoreGCNModel=ResNet-18, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 91.5 | 0 | — | |
| Low Confidence (LC)Model=DenseNet-121, Candidate Model Used?=No, Active Learning Budget=10K2026.05 | 91.37 | 0.16 | 0.9 | |
| High Confidence (HC)Model=ResNet-56, Candidate Model Used?=No2026.05 | 91.3 | 0.26 | 8.33 | |
| BADGEModel=DenseNet-121, Candidate Model Used?=Yes2026.05 | 91.24 | 0 | 7.6 | |
| Beluch et al.Model=DenseNet-121, Candidate Model Used?=Yes, Active Learning Budget=14.5K2026.05 | 91.2 | 0 | — | |
| SuperconvergenceModel=ResNet-56, Candidate Model Used?=No2026.05 | 91.1 | — | — | |
| Uncertainty EstimationModel=DenseNet-121, Candidate Model Used?=Yes2026.05 | 91 | 0 | — | |
| FF-ActiveModel=VGG-16, Candidate Model Used?=Yes2026.05 | 91 | 0 | — | |
| UncertainGCNModel=ResNet-18, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 91 | 0 | — | |
| Yun et al.Model=ResNet-18, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 90.43 | 0 | — | |
| VAALModel=VGG-16, Candidate Model Used?=Yes2026.05 | 90.16 | 0 | — | |
| Low Confidence (LC)Model=ResNet-18, Candidate Model Used?=No, Active Learning Budget=10K2026.05 | 90.12 | 0.25 | 0.6 | |
| Coreset - WSMModel=VGG-16, Candidate Model Used?=Yes2026.05 | 90 | 0 | — | |
| Coreset - FSMModel=VGG-16, Candidate Model Used?=Yes2026.05 | 90 | 0 | — | |
| Adversarial SamplingModel=ResNet-56, Candidate Model Used?=Yes2026.05 | 89 | 0 | — | |
| Interpretability-Aware ViTModel=ViT-Small, Candidate Model Used?=Yes2026.05 | 88.93 | 0 | — | |
| DEALModel=ResNet-18, Candidate Model Used?=Yes2026.05 | 88 | 0 | — | |
| Bootstrapping ViTsModel=ViT-Small, Candidate Model Used?=Yes2026.05 | 87.32 | 0 | — | |
| Low Confidence (LC)Model=SWIN, Candidate Model Used?=No2026.05 | 86.23 | 0.06 | 3.82 | |
| FPGAModel=MobileNetV2, Candidate Model Used?=Yes2026.05 | 86 | 0 | — | |
| Transfer Learning Classifier (TLC)Model=DenseNet-121, Candidate Model Used?=Yes2026.05 | 85.96 | 0 | — | |
| High Confidence and Least Confidence (HCLC)Model=SWIN, Candidate Model Used?=No2026.05 | 85.8 | 0.06 | 3.82 | |
| Low Confidence (LC)Model=VGG-16, Candidate Model Used?=No, Active Learning Budget=20K2026.05 | 84.26 | 0.5 | 3.6 | |
| Discriminative Active Learning (DAL)Model=VGG-16, Candidate Model Used?=Yes, Active Learning Budget=25K2026.05 | 84 | 0 | — | |
| Low Confidence (LC)Model=ViT-Small, Candidate Model Used?=No2026.05 | 83.92 | 0.52 | 3.67 | |
| High Confidence (HC)Model=SWIN, Candidate Model Used?=No2026.05 | 83.88 | 0.06 | 3.82 | |
| High Confidence and Least Confidence (HCLC)Model=ViT-Small, Candidate Model Used?=No2026.05 | 83.81 | 0.52 | 3.67 | |
| LAL-IGradV-VAEModel=CNN (6-layer), Candidate Model Used?=Yes2026.05 | 83.1 | 0 | — | |
| ST-CoNALModel=ResNet-18, Candidate Model Used?=Yes, Active Learning Budget=5K2026.05 | 83.05 | 0 | — | |
| High Confidence (HC)Model=ViT-Small, Candidate Model Used?=No2026.05 | 82.7 | 0.52 | 3.67 | |
| Low Confidence (LC)Model=MobileNetV2, Candidate Model Used?=No, Active Learning Budget=10K2026.05 | 82.53 | 0.52 | 1.5 | |
| Rakesh & JainModel=DenseNet-121, Candidate Model Used?=Yes, Active Learning Budget=8K2026.05 | 82 | 0 | — | |
| BADGEModel=VGG-16, Candidate Model Used?=Yes, Active Learning Budget=40K2026.05 | 82 | 0 | — | |
| PA-CLModel=MobileNetV2, Candidate Model Used?=Yes2026.05 | 82 | 0 | — | |
| Jung et al.Model=VGG-16, Candidate Model Used?=Yes, Active Learning Budget=25K2026.05 | 81.44 | 0 | — | |
| MESAModel=SWIN, Candidate Model Used?=Yes2026.05 | 81.3 | 0 | — | |
| Low Confidence (LC)Model=ResNet-18, Candidate Model Used?=No, Active Learning Budget=5K2026.05 | 81.27 | 0.25 | 0.25 | |
| VAALModel=ResNet-18, Candidate Model Used?=Yes2026.05 | 81 | 0 | — | |
| COPSModel=DenseNet-121, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 80.5 | 0 | — | |
| AOTModel=MobileNetV2, Candidate Model Used?=Yes2026.05 | 80 | 0 | — | |
| Low Confidence (LC)Model=ResNet-56, Candidate Model Used?=No, Active Learning Budget=10K2026.05 | 79.76 | 0.26 | 1.74 | |
| Double-Cross AttacksModel=ResNet-56, Candidate Model Used?=Yes2026.05 | 78.44 | 0 | — | |
| NHTModel=SWIN, Candidate Model Used?=No2026.05 | 78 | — | — | |
| ActiveLossNetModel=ResNet-18, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 77.22 | 0 | — | |
| Triplet ALModel=DenseNet-121, Candidate Model Used?=Yes2026.05 | 76 | 0 | — | |
| COPSModel=ResNet-56, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 74.5 | 0 | — | |
| COPSModel=MobileNetV2, Candidate Model Used?=Yes, Active Learning Budget=10K2026.05 | 74 | 0 | — | |
| Knowledge DistillationModel=ViT-Small, Candidate Model Used?=Yes2026.05 | 73.8 | 0 | — | |
| Gaussian Switch SamplingModel=DenseNet-121, Candidate Model Used?=Yes2026.05 | 72 | 0 | — | |
| BAITModel=DenseNet-121, Candidate Model Used?=Yes2026.05 | 70.91 | 0 | 9.22 | |
| Gaussian Switch SamplingModel=ResNet-18, Candidate Model Used?=Yes2026.05 | 66.11 | 0 | — | |
| BADGEModel=ResNet-18, Candidate Model Used?=Yes, Active Learning Budget=40K2026.05 | 59 | 0 | — |