No Resource, No Benchmarks, No Problem? Evaluating and Improving LLMs for Code Generation in No-Resource Languages

Hugging Face Daily Papers Papers

Summary

This paper tackles code generation for no-resource programming languages by building benchmarks and proposing a method that combines further pre-training with weight difference transfer to create specialized instruction-following models at reduced cost.

Large Language Models (LLMs) have significantly advanced the automation of software engineering tasks. One prominent example is code generation, where an LLM produces code in a specified programming language based on a natural language description. Most research in this area has focused on high-resource languages, such as Python or Java, which benefit from abundant training data. A smaller body of work has explored low-resource languages, which are underrepresented in training corpora. In contrast, no-resource languages for which LLMs have seen virtually no training data remain largely unstudied. These languages often emerge in industry, where organizations develop proprietary or domain-specific languages unsupported by commercial tools like GitHub Copilot. This results in the need for companies to deploy their own in-house code recommenders. To investigate possible solutions in this context, we build and release three code generation benchmarks for no-resource languages, based on two recently proposed programming languages for which very little training data is available. Using these benchmarks, we experiment several solutions to teach LLMs about no-resource languages, including prompt-based techniques as well as pre-training and fine-tuning exploiting the little data available. While further pre-training gives the largest performance gains for no-resource languages, applying it directly to instruction-tuned models harms their ability to follow instructions. To address this, we start from a base model, further pre-training it on the target language, and then inject instruction-following capabilities via weight diff transfer from an instruction model. Such an approach significantly improves code generation capabilities in no-resource settings, allowing companies to cheaply deploy a specialized instruct model without dealing with the computational cost of instruction fine-tuning.
Original Article
View Cached Full Text

Cached at: 06/20/26, 02:27 PM

Paper page - No Resource, No Benchmarks, No Problem? Evaluating and Improving LLMs for Code Generation in No-Resource Languages

Source: https://huggingface.co/papers/2606.16827

Abstract

Research addresses code generation challenges for no-resource programming languages by developing benchmarks and proposing a method that combines further pre-training with weight difference transfer to create specialized instruction-following models at reduced computational cost.

Large Language Models(LLMs) have significantly advanced the automation of software engineering tasks. One prominent example iscode generation, where an LLM produces code in a specified programming language based on a natural language description. Most research in this area has focused on high-resource languages, such as Python or Java, which benefit from abundant training data. A smaller body of work has explored low-resource languages, which are underrepresented in training corpora. In contrast,no-resource languagesfor which LLMs have seen virtually no training data remain largely unstudied. These languages often emerge in industry, where organizations develop proprietary or domain-specific languages unsupported by commercial tools like GitHub Copilot. This results in the need for companies to deploy their own in-house code recommenders. To investigate possible solutions in this context, we build and release threecode generationbenchmarks forno-resource languages, based on two recently proposed programming languages for which very little training data is available. Using these benchmarks, we experiment several solutions to teach LLMs aboutno-resource languages, includingprompt-based techniquesas well aspre-trainingandfine-tuningexploiting the little data available. While furtherpre-traininggives the largest performance gains forno-resource languages, applying it directly toinstruction-tuned modelsharms their ability to follow instructions. To address this, we start from a base model, furtherpre-trainingit on the target language, and then injectinstruction-following capabilitiesviaweight diff transferfrom an instruction model. Such an approach significantly improvescode generationcapabilities in no-resource settings, allowing companies to cheaply deploy a specialized instruct model without dealing with the computational cost of instructionfine-tuning.

View arXiv pageView PDFGitHub1Add to collection

Get this paper in your agent:

hf papers read 2606\.16827

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.16827 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.16827 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.16827 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

@polynoamial: https://x.com/polynoamial/status/2064210146558136827

X AI KOLs Following

This article argues that LLM benchmark performance is increasingly a function of test-time compute, and that current evaluation methods fail to capture capability improvements when controlling for inference budget. It advocates for plotting performance vs. tokens, cost, or time, and discusses implications for safety evaluations.