Research map / Label-efficient and robust learning / Domain adaptation and generalization
Stabilizing In-Context Multi-Source Domain Adaptation for Biomedical Images Through Controls
Ana Sanchez-Fernandez, Thomas Pinetz, Werner Zellinger, Günter Klambauer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add negative controls to the BN adaptation context and meta-training | ARM-BN · label shift alpha=0.01 | accuracy: 0.228 -> 0.894 | Improved |
| add negative controls to the BN adaptation context and meta-training | ARM-BN · label shift alpha=0.1 | accuracy: 0.416 -> 0.906 | Improved |
| add negative controls to the BN adaptation context | ARM-BN · mild shift, new batch (S3) | accuracy: 0.935 -> 0.930 | No clear effect |
| add negative controls and meta-learning | ARM-BN · cross-source S3 -> S8 | accuracy: 0.874 -> 0.884 | Improved |
| add negative controls and meta-learning | ARM-BN · cross-source S8 -> S3 | accuracy: 0.795 -> 0.776 | Hurt |
| add negative controls at test time only (ablation) | ARM-BN · label shift alpha=0.01, equal samples | accuracy: 0.228 -> 0.752 | Improved |
| add control-aware meta-training on top of test-time controls (ablation) | ARM-BN (perturbed + controls) · label shift alpha=0.01 | accuracy: 0.752 -> 0.894 | Improved |
| add meta-learning to control-only BN adaptation (ablation) | BEN · S3 -> S8, alpha=0.01 | accuracy: 0.125 -> 0.663 | Improved |
| add perturbed samples to controls for BN statistics (ablation) | ARM-BEN · S3 -> S8, alpha=0.01 | accuracy: 0.663 -> 0.825 | Improved |
| add entropy minimization on BN affine parameters (ablation) | AdaBN · new batch, alpha=0.01 | accuracy: 0.229 -> 0.156 | Hurt |
| add test-time adaptation (AdaBN) | ResNet50 BN · S3 -> S8 cross-source | accuracy: 0.670 -> 0.838 | Improved |
| add meta-learned in-context adaptation (ARM-BN) to small target batch | AdaBN · smallest labeled batch size | accuracy: 0.130 -> 0.126 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TENT — Tent: Fully Test-Time Adaptation by Entropy Minimization (ICLR 2021) · better on 2 of 2 shared results
- CORAL — CORAL: Disentangling Latent Representations in Long-Tailed Diffusion (NeurIPS 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Test-Time Domain Adaptation by Learning Domain-Aware Batch Normalization (AAAI 2024 · similarity 0.88)
- Taxonomy Adaptive Cross-Domain Adaptation in Medical Imaging via Optimization Trajectory Distillation (ICCV 2023 · similarity 0.88)
- Reducing Divergence in Batch Normalization for Domain Adaptation (AAAI 2025 · similarity 0.87)
- Revisiting the Domain Shift and Sample Uncertainty in Multi-source Active Domain Transfer (CVPR 2024 · similarity 0.85)
- Open-Set Heterogeneous Domain Adaptation: Theoretical Analysis and Algorithm (AAAI 2025 · similarity 0.85)
- Perturbating, Tuning, and Collaborating: Harnessing Vision Foundation Models for Single Domain Generalization on Medical Imaging (AAAI 2025 · similarity 0.85)
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.
