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

Large language models (LLMs) trained on datasets of publicly available sourcecode have established a new state of the art in code generation tasks. However,these models are mostly unaware of the code that exists within a specificproject, preventing the models from making good use of existing APIs. Instead,LLMs often invent, or "hallucinate", non-existent APIs or produce variants ofalready existing code. This paper presents De-Hallucinator, a technique thatgrounds the predictions of an LLM through a novel combination of retrievingsuitable API references and iteratively querying the model with increasinglysuitable context information in the prompt. The approach exploits theobservation that predictions by LLMs often resemble the desired code, but theyfail to correctly refer to already existing APIs. De-Hallucinator automaticallyidentifies project-specific API references related to the model's initialpredictions and adds these references into the prompt. Unlikeretrieval-augmented generation (RAG), our approach uses the initialprediction(s) by the model to iteratively retrieve increasingly suitable APIreferences. Our evaluation applies the approach to two tasks: predicting APIusages in Python and generating tests in JavaScript. We show thatDe-Hallucinator consistently improves the generated code across five LLMs. Inparticular, the approach improves the edit distance by 23.3-50.6% and therecall of correctly predicted API usages by 23.9-61.0% for code completion, andimproves the number of fixed tests that initially failed because ofhallucinations by 63.2%, resulting in a 15.5% increase in statement coveragefor test generation.

De-Hallucinator: Mitigating LLM Hallucinations in Code Generation Tasks via Iterative Grounding