Database Paper Browser

Back to papers

BASE: Bridging the Gap between Cost and Latency for Query Optimization

Summary: BASE: two-stage RL optimizer—train policy on cheap cost signals then transfer a calibrated reward function via an inverse-RL variant to align the policy to latency without costly online execution. Mutual reward–policy refinement improves latency over cost-only learners, cuts training time ≈30% vs SOTA, and generalizes to boost other learned optimizers. (summarized by gpt-5-mini on Feb 09 2026)

Paper ID
13051
Venue
VLDB
Year
2023
Pagerank
4.3950066e-05
Overall Rank
9,108 | 36.64%
DOI
10.14778/3594512.3594525

Incoming Non-self Citations Over Time

Authors

Incoming Citations (Sorted by Pagerank)

Showing 5 of 5 citing papers.

Previous Page 1 / 1 Next

Outgoing Citations (Sorted by Pagerank)

Showing 6 of 6 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank Cited Paper Year Venue Pagerank
71 How Good Are Query Optimizers, Really? 2016 VLDB 0.00059038975
333 Neo: A Learned Query Optimizer 2019 VLDB 0.00027206884
640 Bao: Making Learned Query Optimization Practical 2021 SIGMOD 0.00018759152
2,121 Balsa: Learning a Query Optimizer Without Expert Demonstrations 2022 SIGMOD 9.5017232e-05
3,142 Active Learning for ML Enhanced Database Systems 2020 SIGMOD 7.4815444e-05
4,690 Deploying a Steered Query Optimizer in Production at Microsoft 2022 SIGMOD 5.997226e-05
Previous Page 1 / 1 Next

Semantically Similar Papers