CS PhD student Sonia Cromp uses CHTC to fill in gaps in AI
Observing gaps in Large Language Models’ (LLMs’) abilities to process tabular data motivated UW-Madison Computer Sciences PhD candidate Sonia Cromp to develop Tabby, an architecture modification to LLMs that improves how they handle structured data. Cromp began working on Tabby two and a half years ago, after realizing there were weaknesses in generative modeling, particularly for tabular data, within LLMs.