Research map / Label-efficient and robust learning / Recognition and long-tail
Unifying Knowledge Base Completion with PU Learning to Mitigate the Observation Bias
Jonas Schouterden, Jessa Bekker, Jesse Davis, Hendrik Blockeel
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 inverse propensity weighting instead of an unweighted CWA estimate | Rule confidence estimation on Yago3-10 · SCARp and SAR selection mechanisms | Brier score: 292.5 -> 4.6; 53.8 -> 1.1; 189.5 -> 3.4; 83.8 -> 1.8; 458.6 -> 62.8; 172.7 -> 3.5 | Improved |
| Use inverse propensity weighting rather than the PCA estimator | Rule confidence estimation on Yago3-10 · SARpop (k = 0.01, k = 0.1) | Brier score: 264.2 -> 62.8; 182.7 -> 3.5 | Improved |
| Use inverse propensity weighting rather than the PCA estimator | Rule confidence estimation on Yago3-10 · SARgroup with PCA upheld | Brier score: 22.4 -> 16.9; 12.4 -> 9.1 | Improved |
| Use inverse propensity weighting rather than ICW | Rule confidence estimation on Yago3-10 · SARgroup (cq = 0.5, c¬q = 0.3; cq = 0.5, c¬q = 0.7) | Brier score: 8.1 -> 3.4; 4.0 -> 1.8 | Improved |
| Use noisy propensity scores with −0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SCARp (cp = 0.3) | Brier score: 4.6 -> 154.2 | Hurt |
| Use noisy propensity scores with +0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SCARp (cp = 0.3) | Brier score: 4.6 -> 40.9 | Hurt |
| Use noisy propensity scores with −0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SCARp (cp = 0.7) | Brier score: 1.1 -> 18.1 | Hurt |
| Use noisy propensity scores with +0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SCARp (cp = 0.7) | Brier score: 1.1 -> 10.0 | Hurt |
| Use noisy propensity scores with −0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SARgroup (cq = 0.5, c¬q = 0.3) | Brier score: 3.4 -> 50.3 | Hurt |
| Use noisy propensity scores with +0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SARgroup (cq = 0.5, c¬q = 0.3) | Brier score: 3.4 -> 14.0 | Hurt |
| Use noisy propensity scores with −0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SARgroup (cq = 0.5, c¬q = 0.7) | Brier score: 1.8 -> 6.9 | Hurt |
| Use noisy propensity scores with +0.1 added (ablation) | IPW rule confidence estimator · Yago3-10, SARgroup (cq = 0.5, c¬q = 0.7) | Brier score: 1.8 -> 4.6 | Hurt |
Closest papers by meaning
- Learning from Both Structural and Textual Knowledge for Inductive Knowledge Graph Completion (NeurIPS 2023 · similarity 0.86)
- Deliberation on Priors: Trustworthy Reasoning of Large Language Models on Knowledge Graphs (NeurIPS 2025 · similarity 0.83)
- Do LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awareness (NeurIPS 2025 · similarity 0.83)
- Martingale Score: An Unsupervised Metric for Bayesian Rationality in LLM Reasoning (NeurIPS 2025 · similarity 0.83)
- Can Students Beyond the Teacher? Distilling Knowledge from Teacher’s Bias (AAAI 2025 · similarity 0.83)
- Probing Hidden Knowledge Holes in Unlearned LLMs (NeurIPS 2025 · similarity 0.83)
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.
