In my experience these models are extremely bad at debugging and can often do better either regenerating the code with a tweaked prompt/context/premise. We nee

In my experience these models are extremely bad at debugging and can often do better either regenerating the code with a tweaked prompt/context/premise. We need like a DebugBench to really measure and push the SoTA here since that is a core part of the SWE Journey.