Automatic Semantic Augmentation of Language Model Prompts

Join us for one of our open reading club sessions to discuss a scientific paper from the ICSE '24 conference on "Automatic Semantic Augmentation of Language Model Prompts (for Code Summarization)" alongside one of its authors, Professor Earl Bar from University College London.

It is well known that, among many prompting strategies, few-shot prompting (i.e., adding examples of relevant questions and answers) improves LLM performance on a wide range of tasks.

This paper suggests a method for enhancing automatically selected few-shot examples with semantic facts derived from source code. It also systematically evaluates this strategy on academic benchmarks for two software engineering tasks, code summarization and line completion, over six programming languages.

This event will be held in a hybrid format at the JetBrains Amsterdam Terrace Tower office and will also be streamed online.