Research map / Learning theory and algorithms / Prediction and decision losses
Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Value
Yongchan Kwon, James Zou
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 |
|---|---|---|---|
| Rank each datum by its average score on bootstrap trees that left it out | Data-OOB · 12 classification datasets, mislabeled-data detection | F1-score | Mixed |
| Use Data-OOB values to remove points from lowest to highest | Logistic regression point-removal experiment · pol (n=10000) | test accuracy: 87.9% -> 90.9% | Improved |
| Use Data-OOB values to remove points from lowest to highest | Logistic regression point-removal experiment · jannis (n=10000) | test accuracy: 72.2%; 0.3% degradation | No clear effect |
| Increase the number of random forest trees from 400 to 800 (ablation) | Data-OOB · pol mislabeled-data detection | F1-score: 0.8806 \pm 0.0009 -> 0.8822 \pm 0.0009 | No clear effect |
| Increase the number of random forest trees from 800 to 3200 (ablation) | Data-OOB · pol mislabeled-data detection | F1-score: 0.8822 \pm 0.0009 -> 0.8859 \pm 0.0012 | No clear effect |
| Increase the number of random forest trees from 400 to 3200 (ablation) | Data-OOB · mislabeled-data detection across datasets | F1-score: B\in\{400, 800, 3200\} | No clear effect |
| Use Data-OOB instead of KNN Shapley for point removal | Point-removal experiment · classification datasets | test accuracy | Improved |
| Use Data-OOB instead of AME for point removal | Point-removal experiment · classification datasets | test accuracy | Improved |
Later papers that compared with this method
1 later accepted paper reports DataOob as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Unifying and Optimizing Data Values for Selection via Sequential Decision-Making (ICML 2026 · better on 0 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
- 2D-OOB: Attributing Data Contribution Through Joint Valuation Framework (NeurIPS 2024 · similarity 0.89)
- KAIROS: Scalable Model-Agnostic Data Valuation (NeurIPS 2025 · similarity 0.87)
- One Sample Fits All: Approximating All Probabilistic Values Simultaneously and Efficiently (NeurIPS 2024 · similarity 0.85)
- SAVA: Scalable Learning-Agnostic Data Valuation (ICLR 2025 · similarity 0.85)
- What is Your Data Worth to GPT? LLM-Scale Data Valuation with Influence Functions (NeurIPS 2025 · similarity 0.84)
- Data-Efficient Learning via Clustering-Based Sensitivity Sampling: Foundation Models and Beyond (ICML 2024 · similarity 0.84)
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.
