Opening The Black-Box: Explaining Learned Cost Models For Databases
Summary: First application of AI explainability to learned query cost models: adapted feature-attribution and saliency methods to make deep LCMs interpretable. Demo interactive tool to diagnose tail prediction errors and guide model fixes. (summarized by gpt-5-mini on Feb 09 2026)
Incoming Non-self Citations Over Time
No non-self incoming citations found for this paper in this database.
Authors
- 1. Roman Heinrich (German National Research Center for Information Technology; Technical University of Darmstadt)
- 2. Oleksandr Havrylov (Technical University of Darmstadt)
- 3. Manisha Luthra (German National Research Center for Information Technology; Technical University of Darmstadt)
- 4. Johannes Wehrstein (Technical University of Darmstadt)
- 5. Carsten Binnig (German National Research Center for Information Technology; Technical University of Darmstadt)
BibTeX Citation
@article{heinrich_vldb25,
title = {{Opening The Black-Box: Explaining Learned Cost Models For Databases}},
author = {Heinrich, Roman and Havrylov, Oleksandr and Luthra, Manisha and Wehrstein, Johannes and Binnig, Carsten},
journal = {PVLDB},
series = {{VLDB} '25},
volume = {18},
number = {12},
pages = {5255--5258},
doi = {10.14778/3750601.3750645},
url = {https://doi.org/10.14778/3750601.3750645},
year = {2025}
}Incoming Citations (Sorted by Pagerank)
Showing 1 of 1 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 11,065 | Learned Cost Models for Query Optimization: From Batch to Streaming Systems | 2025 | VLDB | 5.093636e-05 |
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 4 of 4 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 84 | Learned Cardinalities: Estimating Correlated Joins with Deep Learning | 2019 | CIDR | 0.00035838391 |
| 563 | Plan-Structured Deep Neural Network Models for Query Performance Prediction | 2019 | VLDB | 0.0001650812 |
| 2,844 | Zero-Shot Cost Models for Out-of-the-box Learned Cost Prediction | 2022 | VLDB | 8.0608767e-05 |
| 6,088 | How Good are Learned Cost Models, Really? Insights from Query Optimization Tasks | 2025 | SIGMOD | 5.9813965e-05 |
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 10,716 | Demo of LearnedWMP: Workload Memory Prediction Using Deep Query Template Representations | 2025 | SIGMOD |
| 2 | 4,468 | One Model to Rule them All: Towards Zero-Shot Learning for Databases | 2022 | CIDR |
| 3 | 10,825 | Explaining Black-Box Clustering Pipelines With Cluster-Explorer | 2025 | VLDB |
| 4 | 2,195 | Explaining Query Answers with Explanation-Ready Databases | 2016 | VLDB |
| 5 | 5,054 | Explainable AI: Foundations, Applications, Opportunities for Data Management Research | 2022 | SIGMOD |
| 6 | 6,884 | Explaining Inference Queries with Bayesian Optimization | 2021 | VLDB |
| 7 | 2,844 | Zero-Shot Cost Models for Out-of-the-box Learned Cost Prediction | 2022 | VLDB |
| 8 | 11,065 | Learned Cost Models for Query Optimization: From Batch to Streaming Systems | 2025 | VLDB |
| 9 | 6,921 | Rethinking Learned Cost Models: Why Start from Scratch? | 2023 | SIGMOD |
| 10 | 6,088 | How Good are Learned Cost Models, Really? Insights from Query Optimization Tasks | 2025 | SIGMOD |