Do LLMs dream of Type Inference?

Large Language Models (LLMs) have been rapidly evolving in recent years. They offer a completely different approach to challenges that were difficult to solve with classical algorithms.

In this session, I will describe how RBS Goose, an LLM-based type inference tool I am currently developing, works and its current evaluation. For future improvement of RBS Goose, a more detailed evaluation mechanism is needed. In this regard, we will share ideas about TypeEvalRb, a type inference benchmark, by referring to examples from previous studies.