Research map / Label-efficient and robust learning / Test-time adaptation
Robust Calibration with Multi-domain Temperature Scaling
Yaodong Yu, Stephen Bates, Yi Ma, Michael Jordan
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 |
|---|---|---|---|
| Use multi-domain temperature scaling instead of TS | ResNet-50 · ImageNet-C InD domains | per-domain ECE (%): 5.80±0.10 -> 3.84±0.05 | Improved |
| Use multi-domain temperature scaling instead of TS | ResNet-50 · ImageNet-C OOD domains | per-domain ECE (%): 5.70±0.06 -> 4.55±0.04 | Improved |
| Use multi-domain temperature scaling instead of TS | ResNet-50 · WILDS-RxRx1 InD domains | per-domain ECE (%): 9.83±0.57 -> 2.85±0.17 | Improved |
| Use multi-domain temperature scaling instead of TS | ResNet-50 · WILDS-RxRx1 OOD domains | per-domain ECE (%): 13.78±0.43 -> 5.25±0.11 | Improved |
| Use multi-domain temperature scaling instead of TS | ResNet-50 · GLDv2 InD domains | per-domain ECE (%): 11.61±0.09 -> 9.90±0.06 | Improved |
| Use multi-domain temperature scaling instead of TS | ResNet-50 · GLDv2 OOD domains | per-domain ECE (%): 10.75±0.14 -> 9.76±0.12 | Improved |
| Replace OLS with Ridge regression (ablation) | MD-TS with ResNet-50 · WILDS-RxRx1 InD domains | per-domain ECE (%): 2.85 -> 2.88 | No clear effect |
| Replace OLS with Ridge regression (ablation) | MD-TS with ResNet-50 · WILDS-RxRx1 OOD domains | per-domain ECE (%): 5.25 -> 5.26 | No clear effect |
| Replace OLS with Huber regression (ablation) | MD-TS with ResNet-50 · WILDS-RxRx1 InD domains | per-domain ECE (%): 2.85 -> 2.90 | No clear effect |
| Replace OLS with Huber regression (ablation) | MD-TS with ResNet-50 · WILDS-RxRx1 OOD domains | per-domain ECE (%): 5.25 -> 5.29 | No clear effect |
| Replace OLS with kernel ridge regression (ablation) | MD-TS with ResNet-50 · WILDS-RxRx1 InD domains | per-domain ECE (%): 2.85 -> 2.85 | No clear effect |
| Replace OLS with kernel ridge regression (ablation) | MD-TS with ResNet-50 · WILDS-RxRx1 OOD domains | per-domain ECE (%): 5.25 -> 4.99 | Improved |
Closest papers by meaning
- Sample-Dependent Adaptive Temperature Scaling for Improved Calibration (AAAI 2023 · similarity 0.89)
- Consistency-Guided Temperature Scaling Using Style and Content Information for Out-of-Domain Calibration (AAAI 2024 · similarity 0.89)
- Pseudo-Calibration: Improving Predictive Uncertainty Estimation in Unsupervised Domain Adaptation (ICML 2024 · similarity 0.86)
- On the Limitations of Temperature Scaling for Distributions with Overlaps (ICLR 2024 · similarity 0.86)
- Parametric ρ-Norm Scaling Calibration (AAAI 2025 · similarity 0.86)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 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.
