Research map / Label-efficient and robust learning / Recognition and long-tail
Auxiliary Training: Towards Accurate and Robust Models
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Train clean samples with a primary classifier and corrupted samples with corruption-specific auxiliary classifiers, then merge their weights | ResNet18 · CIFAR100 | accuracy: 77.09 -> 79.47 | Helped |
| Use auxiliary training with corruption-specific classifiers | ResNet18 · CIFAR100-C | corruption error (CE): 92.21 -> 69.34 | Helped |
| Add Selective BN (ablation) | Auxiliary Training · CIFAR100 | accuracy: 76.37 -> 79.47 | Helped |
| Add Selective BN (ablation) | Auxiliary Training · CIFAR100-C | corruption error (CE): 69.52 -> 69.34 | Helped |
| Add input-aware self distillation (ablation) | Auxiliary Training · CIFAR100 | accuracy: 78.44 -> 79.47 | Helped |
| Add input-aware self distillation (ablation) | Auxiliary Training · CIFAR100-C | corruption error (CE): 73.67 -> 69.34 | Helped |
| Add attention modules to auxiliary classifiers (ablation) | Auxiliary Training · CIFAR100 | accuracy: 77.50 -> 79.47 | Helped |
| Add attention modules to auxiliary classifiers (ablation) | Auxiliary Training · CIFAR100-C | corruption error (CE): 70.79 -> 69.34 | Helped |
| Add weight merging between primary and auxiliary classifiers (ablation) | Auxiliary Training · CIFAR100 | accuracy: 78.32 -> 79.47 | Helped |
| Add weight merging between primary and auxiliary classifiers (ablation) | Auxiliary Training · CIFAR100-C | corruption error (CE): 70.43 -> 69.34 | Helped |
Closest papers by meaning
- DiffAug: A Diffuse-and-Denoise Augmentation for Training Robust Classifiers (NeurIPS 2024 · similarity 0.85)
- Foundation Model-oriented Robustness: Robust Image Model Evaluation with Pretrained Models (ICLR 2024 · similarity 0.85)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.84)
- Dynamic Sparse Training versus Dense Training: The Unexpected Winner in Image Corruption Robustness (ICLR 2025 · similarity 0.84)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.83)
- Leveraging Normalization Layer in Adapters with Progressive Learning and Adaptive Distillation for Cross-Domain Few-Shot Learning (AAAI 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.