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/

Zendy Plus

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

Zendy Tools

Zendy Open

Language Models (LMs) have become widely used in software engineering,especially for tasks such as code generation, where they are referred to ascode LMs. These models have proven effective in generating code, making iteasier for developers to automate coding activities. However, research hashighlighted a significant limitation: despite their effectiveness, LMs oftenproduce code that is incorrect, buggy, or not fully functional. Updating thesemodels with limited data can be prohibitively challenging, yet it is essentialto maximize their utility. This may require hot-fix techniques (updating modelswith limited data) to resolve. In this paper, we propose \ul{M}odel\ul{I}mprovement via \ul{N}euron \ul{T}argeting (\textsc{MINT}), a novelapproach for repairing code LMs. MINT leverages the semantic property oflanguage models to perform neuron-level repairs in a novel way. Further, byanalyzing the relationships between the model's latent representations, theincorrect outputs, and the desired outputs, \textsc{MINT} determines whichneurons are worth updating. This approach ensures that only the neurons crucialto the model's failure are targeted, avoiding unnecessary changes and allowingfor a more efficient and precise repair process. \textsc{MINT} is effective,efficient, and reliable, capable of correcting a neural model by patching aminimum number of neurons (usually one or two neurons). Our approach isevaluated on three coding tasks: line-level code generation, shellcodegeneration, and intent-to-bash translation. The experimental resultsdemonstrate that the proposed approach significantly outperforms thestate-of-the-art in both effectiveness and efficiency measures. In addition, weanalyze and discuss the side effects of model repair techniques, including thebalance between generalization and specificity, and the performance aftermultiple repairs in succession.

Neuron Patching: Semantic-based Neuron-level Language Model Repair for Code Generation