Researchers have found that large language models (LLMs) often misuse APIs when generating Java code. The study tested four LLMs, including GPT-3.5 and GPT-4 from OpenAI, and found high rates of API misuse. The researchers argue that while LLMs have improved in code generation, the reliability and robustness of the code in real-world production remains a significant issue, indicating a large scope for improvement.
Read more…
