A Local Agentic AI Framework to Generate Edge Case Scenarios for Safety Assurance
arXiv:2610.13479v1 Announce Type: new Abstract: Generating edge case scenarios is crucial for training, testing and safety assuring autonomous systems. This identifies hazardous failures to increase system reliability and robustness, ensures the system can still function safely on the boundaries of its operational design domain and learns the expected system behaviour for these edge cases. The rapid advancement in the generation and analysis capabilities of language models (LMs) has expanded th…