[papichulo4]

Totally agree, seen it too. Do you think it can be fixed over time with better training data and optimization? Or, is this a fundamental limitation that LLMs will never overcome?