Research map / Label-efficient and robust learning / Domain adaptation and generalization
Variational Model Perturbation for Source-Free Domain Adaptation
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 |
|---|---|---|---|
| add variational model perturbation to SHOT | SHOT · Office offline SFDA mean | accuracy (%): 88.6 -> 90.0 | Helped |
| add variational model perturbation to SHOT | SHOT · Office-Home offline SFDA mean | accuracy (%): 71.8 -> 72.8 | Helped |
| add variational model perturbation to SHOT | SHOT · Office-Home generalized SFDA | harmonic mean accuracy (%): 71.0 -> 74.0 | Helped |
| SHOT fine-tuning of the source model | source model · Office-Home generalized SFDA source accuracy | accuracy (%): 83.6 -> 71.7 | Hurt |
| perturb Tent (variational model perturbation) vs Tent | Tent · CIFAR10-C continual online SFDA | test error (%): +2.8% advantage over Tent | Helped |
| update all weights (Tent-FT) instead of BN only | Tent · CIFAR10-C continual online SFDA | test error (%): 0.9% improvement | Helped |
| perturbation vs BN modulation and fine-tuning (ablation) | adaptation scheme · Office-Home and CIFAR100-C | accuracy / error | Helped |
| adaptive per-kernel prior instead of isotropic Gaussian prior (ablation) | variational model perturbation · Office-Home | accuracy | Helped |
| parameter sharing within conv kernels (ablation) | variational model perturbation · Office / Office-Home / CIFAR10-C | accuracy: 24.03 million -> 600 thousand params | No clear effect |
| perturb the last few layers vs leave them unperturbed (ablation) | variational model perturbation · ablation over layers | accuracy | Helped |
Closest papers by meaning
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.89)
- Diffusion-Driven Progressive Target Manipulation for Source-Free Domain Adaptation (NeurIPS 2025 · similarity 0.88)
- Proxy Denoising for Source-Free Domain Adaptation (ICLR 2025 · similarity 0.88)
- Revisiting the Domain Shift and Sample Uncertainty in Multi-source Active Domain Transfer (CVPR 2024 · similarity 0.87)
- Efficient Source-Free Time-Series Adaptation via Parameter Subspace Disentanglement (ICLR 2025 · similarity 0.87)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 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.