arXiv:2602.20213v2 Announce Type: replace-cross Abstract: The evaluation of Large Language Models (LLMs) for code generation relies heavily on the quality and robustness of test cases. However, existing benchmarks often lack coverage for subtle corner cases, allowing incorrect solutions to pass. To bridge this gap, we propose CodeHacker, an...
Läs hela artikeln hos källan.
Kommentarer (0)
Inga kommentarer ännu. Bli först med att kommentera!