08.12.2025Database Theory Test of Time Award

The paper «Declarative Probabilistic Programming with Datalog» has best met the «test of time» and had the highest impact in terms of research, methodology, conceptual contribution, or transfer to practice over the past decade. 

Citation from the ICDT 2026 Test-of-Time committee:

“The paper proposes Generative Datalog, a probabilistic extension of Datalog that allows sampling from discrete probability distributions. Generative Datalog can be seen as a declarative probabilistic programming language that operates on standard relational databases. The idea is simple but elegant: Given that we can view an existential Datalog program as a generator of families of models, why not turn it into a generator of a probabilistic model? On the side of language design, essentially all it takes is to attach probability distributions to the tuple-generating dependencies. The easy concept of the language is paired with a surprisingly deep mathematical background: Even though all distributions discussed in this initial paper are discrete, laying the semantic groundwork already requires excursions into measure theory. The paper explains the language, defines the semantics, a probabilistic version of the chase, discusses adding constraints in the spirit of probabilistic programming, and touches upon the equivalence problem for programs. It generated a significant amount of follow-up in a variety of top venues spanning database theory, database systems, and programming languages.”

Link to the extended journal version of the paper: https://dl.acm.org/doi/10.1145/3132700
Link to the citation: https://edbticdt2026.github.io/?contents=ICDT_Test-of-time.html