[tlonny]

Given the size of the training data - I don’t think it would economical to validate all training data with high-end LLM models.

[sedatk]

True. Maybe it can be dumbed down to a low-end model specifically for this type of detection.