Research map / Label-efficient and robust learning / Test-time adaptation
Adaptive Risk Minimization: Learning to Adapt to Domain Shift
Read the paperCheck 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 |
|---|---|---|---|
| meta-train context network on single-domain batches (ARM-CML vs context ablation) (ablation) | ARM-CML · Rotated MNIST | worst-case accuracy: 63.5 -> 88.0 | Helped |
| sample training batches from a single domain for learned loss (ARM-LL vs LL ablation) (ablation) | ARM-LL · Rotated MNIST | worst-case accuracy: 79.9 -> 88.9 | Helped |
| add domain meta-training to BN adaptation (ARM-BN) (ablation) | BN adaptation · Tiny ImageNet-C | worst-case accuracy: 26.5 -> 28.3 | Helped |
| add domain meta-training to BN adaptation (ARM-BN) (ablation) | BN adaptation · CIFAR-10-C | worst-case accuracy: 60.6 -> 61.7 | Helped |
| add domain meta-training to learned loss on Tiny ImageNet-C (ablation) | ARM-LL · Tiny ImageNet-C | worst-case accuracy: 21.6 -> 25.4 | Helped |
| use probabilistic context with Gaussian prior instead of deterministic context (ablation) | ARM-CML · Rotated MNIST | worst-case accuracy: 88.0 -> 82.6 | Hurt |
| use probabilistic context with Gaussian prior instead of deterministic context (ablation) | ARM-CML · FEMNIST | worst-case accuracy: 70.9 -> 65.1 | Hurt |
| learn domains with a VAE rather than use given domains (ablation) | ARM-CML · Rotated MNIST | worst-case accuracy: 88.0 -> 81.7 | Hurt |
| ARM-CML vs ERM using given domains | ERM · Rotated MNIST | worst-case accuracy: 74.5 -> 88.0 | Helped |
| ARM-BN vs ERM | ERM · RxRx1 | accuracy: 70.3 -> 87.2 | Helped |
| ARM-BN vs ERM | ERM · FMoW | worst-case accuracy: 32.3 -> 24.6 | Hurt |
| ARM-BN vs ERM | ERM · iWildCam | accuracy: 71.6 -> 70.3 | No clear effect |
Closest papers by meaning
- Estimating Generalization under Distribution Shifts via Domain-Invariant Representations (ICML 2020 · similarity 0.91)
- Robust Domain Generalization under Divergent Marginal and Conditional Distributions (NeurIPS 2026 · similarity 0.88)
- Test-Time Domain Adaptation by Learning Domain-Aware Batch Normalization (AAAI 2024 · similarity 0.88)
- Domain-wise Data Acquisition to Improve Performance under Distribution Shift (ICML 2024 · similarity 0.87)
- Optimal Aggregation of Prediction Intervals under Unsupervised Domain Shift (NeurIPS 2024 · similarity 0.87)
- Lost Domain Generalization Is a Natural Consequence of Lack of Training Domains (AAAI 2024 · similarity 0.87)
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.