@inproceedings{kumar-etal-2026-datakernelbench,
title = "{D}ata{K}ernel{B}ench: Can {LLM}s Optimize Database Queries on {GPU}s?",
author = "Kumar, Gokul Karthik and
Perlitz, Yotam and
Lammie, Corey and
Giovannini, Andrea and
Hose, Katja",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.436/",
pages = "9153--9168",
ISBN = "979-8-89176-462-0",
abstract = "GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels. Existing LLM kernel benchmarks focus on machine learning operators, leaving irregular, heterogeneous, data-movement-heavy database-style operators untested. We introduce **DataKernelBench**, which translates SQL into validated PyTorch **TorchPlan** programs and evaluates LLMs that optimize either the core tensor-bounded snippet or the full query in CUDA or Triton through execution-guided repair. Across ten proprietary and open-weight models on **TPC-H SF10** with an H100 GPU, the strongest full-query CUDA configuration achieves **2.11{\texttimes} speedup over `torch.compile{`}** at full pass rate. We find that higher-performing implementations commonly use kernel fusion and execution-strategy changes, stronger models benefit most from full-query specialization, and workload context matters more than hardware context. To handle data larger than GPU memory, we extend TorchPlan with Dask-cuDF for on-demand partition loading on **TPC-H SF100** with four H100 GPUs, achieving **2.54{\texttimes}** speedup.\par **Project page:** https://kerneldf.github.io/datakernelbench"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kumar-etal-2026-datakernelbench">
<titleInfo>
<title>DataKernelBench: Can LLMs Optimize Database Queries on GPUs?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Gokul</namePart>
<namePart type="given">Karthik</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yotam</namePart>
<namePart type="family">Perlitz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Corey</namePart>
<namePart type="family">Lammie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrea</namePart>
<namePart type="family">Giovannini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Katja</namePart>
<namePart type="family">Hose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels. Existing LLM kernel benchmarks focus on machine learning operators, leaving irregular, heterogeneous, data-movement-heavy database-style operators untested. We introduce **DataKernelBench**, which translates SQL into validated PyTorch **TorchPlan** programs and evaluates LLMs that optimize either the core tensor-bounded snippet or the full query in CUDA or Triton through execution-guided repair. Across ten proprietary and open-weight models on **TPC-H SF10** with an H100 GPU, the strongest full-query CUDA configuration achieves **2.11× speedup over ‘torch.compile‘** at full pass rate. We find that higher-performing implementations commonly use kernel fusion and execution-strategy changes, stronger models benefit most from full-query specialization, and workload context matters more than hardware context. To handle data larger than GPU memory, we extend TorchPlan with Dask-cuDF for on-demand partition loading on **TPC-H SF100** with four H100 GPUs, achieving **2.54×** speedup.\par **Project page:** https://kerneldf.github.io/datakernelbench</abstract>
<identifier type="citekey">kumar-etal-2026-datakernelbench</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.436/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>9153</start>
<end>9168</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T DataKernelBench: Can LLMs Optimize Database Queries on GPUs?
%A Kumar, Gokul Karthik
%A Perlitz, Yotam
%A Lammie, Corey
%A Giovannini, Andrea
%A Hose, Katja
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F kumar-etal-2026-datakernelbench
%X GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels. Existing LLM kernel benchmarks focus on machine learning operators, leaving irregular, heterogeneous, data-movement-heavy database-style operators untested. We introduce **DataKernelBench**, which translates SQL into validated PyTorch **TorchPlan** programs and evaluates LLMs that optimize either the core tensor-bounded snippet or the full query in CUDA or Triton through execution-guided repair. Across ten proprietary and open-weight models on **TPC-H SF10** with an H100 GPU, the strongest full-query CUDA configuration achieves **2.11× speedup over ‘torch.compile‘** at full pass rate. We find that higher-performing implementations commonly use kernel fusion and execution-strategy changes, stronger models benefit most from full-query specialization, and workload context matters more than hardware context. To handle data larger than GPU memory, we extend TorchPlan with Dask-cuDF for on-demand partition loading on **TPC-H SF100** with four H100 GPUs, achieving **2.54×** speedup.\par **Project page:** https://kerneldf.github.io/datakernelbench
%U https://aclanthology.org/2026.findings-emnlp.436/
%P 9153-9168
Markdown (Informal)
[DataKernelBench: Can LLMs Optimize Database Queries on GPUs?](https://aclanthology.org/2026.findings-emnlp.436/) (Kumar et al., Findings 2026)
ACL
- Gokul Karthik Kumar, Yotam Perlitz, Corey Lammie, Andrea Giovannini, and Katja Hose. 2026. DataKernelBench: Can LLMs Optimize Database Queries on GPUs?. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 9153–9168, Budapest, Hungary. Association for Computational Linguistics.