@inproceedings{zarb-etal-2026-minimal,
title = "Minimal Grammars as Supervision for {DSL} Generation",
author = "Zarb, David Vella and
Parks, Geoff and
Kipouros, Timoleon",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1435/",
pages = "32782--32802",
ISBN = "979-8-89176-459-0",
abstract = "Language models approach saturation on general-purpose code benchmarks but routinely fail on domain-specific languages (DSLs), where paired training data is scarce and regularities learned from mainstream languages transfer poorly. We propose augmenting each training example with a minimal grammar: the subset of DSL productions needed to derive the target program, typically 2{--}20{\%} of the full grammar. This provides a strong structural prior, and we fine-tune the model to generate the program conditioned on the query and grammar. We predict grammars at inference via retrieval: a frontier model is prompted with (query, grammar) pairs, annotated with a chain-of-thought trace. Across five benchmarks spanning semantic parsing (SMCalFlow, GeoQuery, Overnight) and hardware description languages (HDLs) (Verilog, SPICE), our method improves over grammar-free fine-tuning on three open-weight backbones. Gains reach 36.8pp of program accuracy on SMCalFlow, with consistent improvements on HDLs. Applied to in-context learning, our predicted grammars raise GPT-5.5 accuracy on SMCalFlow from 51.4{\%} (grammar prompting) to 94.3{\%}, with similar gains for Claude Opus 4.7. Ablations rule out the extra training tokens as the source of these gains."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zarb-etal-2026-minimal">
<titleInfo>
<title>Minimal Grammars as Supervision for DSL Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="given">Vella</namePart>
<namePart type="family">Zarb</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Geoff</namePart>
<namePart type="family">Parks</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Timoleon</namePart>
<namePart type="family">Kipouros</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Language models approach saturation on general-purpose code benchmarks but routinely fail on domain-specific languages (DSLs), where paired training data is scarce and regularities learned from mainstream languages transfer poorly. We propose augmenting each training example with a minimal grammar: the subset of DSL productions needed to derive the target program, typically 2–20% of the full grammar. This provides a strong structural prior, and we fine-tune the model to generate the program conditioned on the query and grammar. We predict grammars at inference via retrieval: a frontier model is prompted with (query, grammar) pairs, annotated with a chain-of-thought trace. Across five benchmarks spanning semantic parsing (SMCalFlow, GeoQuery, Overnight) and hardware description languages (HDLs) (Verilog, SPICE), our method improves over grammar-free fine-tuning on three open-weight backbones. Gains reach 36.8pp of program accuracy on SMCalFlow, with consistent improvements on HDLs. Applied to in-context learning, our predicted grammars raise GPT-5.5 accuracy on SMCalFlow from 51.4% (grammar prompting) to 94.3%, with similar gains for Claude Opus 4.7. Ablations rule out the extra training tokens as the source of these gains.</abstract>
<identifier type="citekey">zarb-etal-2026-minimal</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1435/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>32782</start>
<end>32802</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Minimal Grammars as Supervision for DSL Generation
%A Zarb, David Vella
%A Parks, Geoff
%A Kipouros, Timoleon
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F zarb-etal-2026-minimal
%X Language models approach saturation on general-purpose code benchmarks but routinely fail on domain-specific languages (DSLs), where paired training data is scarce and regularities learned from mainstream languages transfer poorly. We propose augmenting each training example with a minimal grammar: the subset of DSL productions needed to derive the target program, typically 2–20% of the full grammar. This provides a strong structural prior, and we fine-tune the model to generate the program conditioned on the query and grammar. We predict grammars at inference via retrieval: a frontier model is prompted with (query, grammar) pairs, annotated with a chain-of-thought trace. Across five benchmarks spanning semantic parsing (SMCalFlow, GeoQuery, Overnight) and hardware description languages (HDLs) (Verilog, SPICE), our method improves over grammar-free fine-tuning on three open-weight backbones. Gains reach 36.8pp of program accuracy on SMCalFlow, with consistent improvements on HDLs. Applied to in-context learning, our predicted grammars raise GPT-5.5 accuracy on SMCalFlow from 51.4% (grammar prompting) to 94.3%, with similar gains for Claude Opus 4.7. Ablations rule out the extra training tokens as the source of these gains.
%U https://aclanthology.org/2026.emnlp-main.1435/
%P 32782-32802
Markdown (Informal)
[Minimal Grammars as Supervision for DSL Generation](https://aclanthology.org/2026.emnlp-main.1435/) (Zarb et al., EMNLP 2026)
ACL
- David Vella Zarb, Geoff Parks, and Timoleon Kipouros. 2026. Minimal Grammars as Supervision for DSL Generation. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 32782–32802, Budapest, Hungary. Association for Computational Linguistics.