Training Dataset Validation & Benchmark Testing

[mrehanali]

Allow users to define a training dataset of pre-validated question-answer pairs with expected confidence scores. The validator can then run batch validation against this dataset to benchmark performance, detect regressions, and ensure consistency. This enables developers to test their AI systems against known good/bad examples and track validation accuracy over time. Useful for quality assurance, A/B testing different models, and maintaining validation standards.