This work aims to advance sound event detection (SED) research by presentinga new large language model (LLM)-powered dataset namely wild domesticenvironment sound event detection (WildDESED). It is crafted as an extension tothe original DESED dataset to reflect diverse acoustic variability and complexnoises in home settings. We leveraged LLMs to generate eight different domesticscenarios based on target sound categories of the DESED dataset. Then weenriched the scenarios with a carefully tailored mixture of noises selectedfrom AudioSet and ensured no overlap with target sound. We consider widelypopular convolutional neural recurrent network to study WildDESED dataset,which depicts its challenging nature. We then apply curriculum learning bygradually increasing noise complexity to enhance the model's generalizationcapabilities across various noise levels. Our results with this approach showimprovements within the noisy environment, validating the effectiveness on theWildDESED dataset promoting noise-robust SED advancements.