English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Natural language to code generation is an important application area of LLMsand has received wide attention from the community. The majority of relevantstudies have exclusively concentrated on increasing the quantity and functionalcorrectness of training sets while disregarding other stylistic elements ofprograms. More recently, data quality has garnered a lot of interest andmultiple works have showcased its importance for improving performance. In thiswork, we investigate data quality for code and find that making the code morestructured and readable leads to improved code generation performance of thesystem. We build a novel data-cleaning pipeline that uses these principles totransform existing programs by 1.) renaming variables, 2.) modularizing anddecomposing complex code into smaller helper sub-functions, and 3.) insertingnatural-language based plans via LLM based transformations. We evaluate ourapproach on two challenging algorithmic code generation benchmarks and findthat fine-tuning CodeLLaMa-7B on our transformed modularized programs improvesthe performance by up to 30% compared to fine-tuning on the original dataset.Additionally, we demonstrate improved performance from using a smaller amountof higher-quality data, finding that a model fine-tuned on the entire originaldataset is outperformed by a model trained on 15% of our cleaned dataset. Evenin comparison to closed-source models, our models outperform the much largerAlphaCoder models.

LLM-Assisted Code Cleaning For Training Accurate Code Generators