Balsa: Learning a Query Optimizer Without Expert Demonstrations
Summary: Balsa uses deep RL to learn a query optimizer without expert demonstrations, from a simple simulator to safe real-execution fine-tuning. On Join Order Benchmark, it matches two expert optimizers after ~2 hours and beats them up to 2.8x with training. (summarized by gpt-5-nano on Feb 09 2026)
Incoming Non-self Citations Over Time
Authors
- 1. Zongheng Yang (University of California Berkeley)
- 2. Wei-Lin Chiang (University of California Berkeley)
- 3. Sifei Luan (University of California Berkeley)
- 4. Gautam Mittal (University of California Berkeley)
- 5. Michael Luo (University of California Berkeley)
- 6. Ion Stoica (University of California Berkeley)
BibTeX Citation
@inproceedings{yang_sigmod22,
title = {{Balsa: Learning a Query Optimizer Without Expert Demonstrations}},
author = {Yang, Zongheng and Chiang, Wei-Lin and Luan, Sifei and Mittal, Gautam and Luo, Michael and Stoica, Ion},
series = {{SIGMOD} '22},
booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
publisher = {Association for Computing Machinery},
doi = {10.1145/3514221.3517885},
url = {https://dl.acm.org/doi/10.1145/3514221.3517885},
year = {2022}
}
Incoming Citations (Sorted by Pagerank)
Showing 9 of 59 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 10,741 | OBELISK: Efficient Offline Query Planning with Bayesian Optimization-Informed Language Model Reasoning | 2026 | VLDB | 4.9793485e-05 |
| 10,751 | Toward Drift-Aware Database Benchmarking | 2026 | VLDB | 4.9793485e-05 |
| 10,883 | QDBO: A Real-time Quantum-augmented Database System Optimizer | 2026 | VLDB | 4.9793485e-05 |
| 10,924 | TATA: An Efficient Framework for Task Transfer in Query Plan Representation | 2026 | VLDB | 4.9793485e-05 |
| 10,938 | Towards Industrial-Scale Parametric Query Optimization | 2026 | VLDB | 4.9793485e-05 |
| 11,283 | Robust Plan Evaluation based on Approximate Probabilistic Machine Learning | 2025 | VLDB | 4.9793485e-05 |
| 11,425 | Learned Cost Models for Query Optimization: From Batch to Streaming Systems | 2025 | VLDB | 4.9793485e-05 |
| 11,443 | LEAP: A Low-cost Spark SQL Query Optimizer using Pairwise Comparison | 2025 | VLDB | 4.9793485e-05 |
| 11,453 | RankPQO: Learning-to-Rank for Parametric Query Optimization | 2025 | VLDB | 4.9793485e-05 |
Outgoing Citations (Sorted by Pagerank)
Showing 16 of 16 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 6,308 | Is Your Learned Query Optimizer Behaving As You Expect? A Machine Learning Perspective | 2024 | VLDB |
| 2 | 5,110 | Steering Query Optimizers: A Practical Take on Big Data Workloads | 2021 | SIGMOD |
| 3 | 4,482 | LearnedSQLGen: Constraint-aware SQL Generation using Reinforcement Learning | 2022 | SIGMOD |
| 4 | 3,487 | LOGER: A Learned Optimizer towards Generating Efficient and Robust Query Execution Plans | 2023 | VLDB |
| 5 | 9,954 | Conformal Prediction for Verifiable Learned Query Optimization | 2025 | VLDB |
| 6 | 4,683 | AutoSteer: Learned Query Optimization for Any SQL Database | 2023 | VLDB |
| 7 | 1,800 | SkinnerDB: Regret-Bounded Query Evaluation via Reinforcement Learning | 2018 | VLDB |
| 8 | 9,276 | BASE: Bridging the Gap between Cost and Latency for Query Optimization | 2023 | VLDB |
| 9 | 362 | Bao: Making Learned Query Optimization Practical | 2021 | SIGMOD |
| 10 | 3,060 | Towards a Hands-Free Query Optimizer through Deep Learning | 2019 | CIDR |