Comment by sedatk Comment by sedatk a year ago 2 replies Copy Link View on Hacker News Both ChatGPT 4o and Claude 3.5 Sonnet can identify the generated page content as "random words".
Copy Link tlonny a year ago Collapse Comment - Given the size of the training data - I don’t think it would economical to validate all training data with high-end LLM models. Reply View | 1 reply Copy Link sedatk a year ago Parent Collapse Comment - True. Maybe it can be dumbed down to a low-end model specifically for this type of detection. Reply View | 0 replies
Copy Link sedatk a year ago Parent Collapse Comment - True. Maybe it can be dumbed down to a low-end model specifically for this type of detection. Reply View | 0 replies
Given the size of the training data - I don’t think it would economical to validate all training data with high-end LLM models.