From the source
Replit describes its methodology for training a code repair model using LSP diagnostics and synthetic diffs, aiming to create a Replit-native AI model that can fix code errors directly within the development environment.
Building LLMs for Code Repair
Replit's approach to training a code repair model using LSP diagnostics and synthetic diffs.
Research paper
From the source
We set out to identify a scenario where we could develop a model that could also become a useful tool for our current developers and settled on code repair. ... Given the abundance of training data, repairing code errors using LSP diagnostics is therefore the ideal setting to build our first Replit-native AI model.
replit.com
