Evaluating Large Language Models for Code Editing Instructions
The authors introduce CanItEdit, a benchmark to assess instructional code editing skills of Code LLMs. They evaluate various models and propose fine-tuning methods to enhance code editing capabilities.
