4 ms·
Instead of training an LLM to make comments, train the LLM to generate test cases and then guess the line of code that breaks the test.
by lupire 2y ago
Instead of training an LLM to make comments, train the LLM to generate test cases and then guess the line of code that breaks the test.
- dakshgupta 2y agoYou might enjoy @goodside on Twitter. He’s a prompt engineer at Scale AI and a lot of his observations and techniques are fascinating.