Comment by codelion

That's a great point about the limitations of traditional OCR with rotated or poorly scanned documents. I agree that VLMs really shine when it comes to understanding context and extracting information beyond just the text itself. It's pretty cool how they can map implicit relationships, like those X-axis labels you mentioned.