Research map / Label-efficient and robust learning / Domain adaptation and generalization
Ranking Models in Unlabeled New Environments
Xiaoxiao Sun, Yunzhong Hou, Weijian Deng, Hongdong Li, Liang Zheng
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 searched proxy instead of source validation set as proxy | proxy for model ranking · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho: 0.320 -> 0.858 | Improved |
| use PersonX instead of source validation set as proxy | proxy for model ranking · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho: 0.320 -> 0.816 | Improved |
| add camera annotation to the proxy search (ablation) | proxy search · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho: 0.858 -> 0.882 | Improved |
| use searched proxy instead of engine-based synthesis | proxy for model ranking · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho: 0.756 -> 0.858 | Improved |
| combine FID and variance gap in sampling score (lambda 0.6) rather than either alone (ablation) | proxy search · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho | Improved |
| replace FID with MMD in sampling score (ablation) | proxy search · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho: -0.0056 Spearman rho and -0.0161 Kendall tau | No clear effect |
| apply SPGAN style transfer to proxy sets (ablation) | proxy sets · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho | Mixed |
| use intermediate cluster number K=20 rather than too small or too large (ablation) | proxy search · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho | Improved |
| increase number of proxy IDs N (ablation) | proxy search · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho | No clear effect |
| use searched proxy instead of random combination of datasets | proxy for model ranking · Model ranking, MSMT17 source -> DukeMTMC-reID target | Spearman rho: +0.133 of rho and +0.176 of tau | Improved |
| train IDE on searched proxy instead of pseudo-label data | IDE re-ID model · DukeMTMC-reID test, IDE trained on different sets | mAP: 50.4 -> 27.9 | Hurt |
Closest papers by meaning
- Can Small Training Runs Reliably Guide Data Curation? Rethinking Proxy-Model Practice (ICLR 2026 · similarity 0.86)
- ModelLens: Finding the Best for Your Task from Myriads of Models (NeurIPS 2026 · similarity 0.81)
- DD-Ranking: Rethinking the Evaluation of Dataset Distillation (NeurIPS 2026 · similarity 0.80)
- Ranked from Within: Ranking Large Multimodal Models Without Labels (ICML 2025 · similarity 0.80)
- Evaluating multiple models using labeled and unlabeled data (NeurIPS 2025 · similarity 0.80)
- NEAR: A Training-Free Pre-Estimator of Machine Learning Model Performance (ICLR 2025 · similarity 0.80)
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.
