Mistral OCR • Hacker News Board

High accuracy is the goal! But the multimodal approach introduces some complexities that can impact real-world performance. We break it down in our review: https://undatas.io/blog/posts/in-depth-review-of-mistral-ocr... As for use cases, it really depends on how well it handles edge cases…

Re: Mistral OCR

#447

Post by utkarshphirke » Thu, Mar 13, 2025, 9:30 AM UTC

raffraffraff wrote: post #380

Forgive my absolute ignorance, I should probably run this through a chat bot before posting ... So I'm updating my post with answers now! Q: Do LLMs specialise in "document level" recognition based on headings, paragraphs, columns tables etc? Ie: ignore words and characters for now and attempt to recognise a known document format. A: Not most LLMs, but those with multimodal / vision capability could (eg DeepSeek Visi…

LLMs are quite poor at rating their own confidence. Your best bet is to train a task specific LLM and ensure it is not overfit

We benchmarked it here - https://news.ycombinator.com/item?id=43350816

Re: Mistral OCR

#448

Post by utkarshphirke » Thu, Mar 13, 2025, 9:32 AM UTC

deadbabe wrote: post #51

LLM based OCR is a disaster, great potential for hallucinations and no estimate of confidence. Results might seem promising but you’ll always be wondering.

Absolutely right - we tried estimating LLM confidence and the results are not great. Any process that requires reliability will struggle with LLM OCR.

https://news.ycombinator.com/item?id=43350816

Re: Mistral OCR

#449

Post by ronald263 » Thu, Mar 13, 2025, 7:11 PM UTC

[dead]

Re: Mistral OCR

#450

Post by nnurmanov » Sat, Mar 15, 2025, 6:31 AM UTC

infecto wrote: post #85

kapitalx wrote: post #81

Co-founder of doctly.ai here (OCR tool) I love mistral and what they do. I got really excited about this, but a little disappointed after my first few tests. I tried a complex table that we use as a first test of any new model, and Mistral OCR decided the entire table should just be extracted as an 'image' and returned this markdown: ``` ![img-0.jpeg](img-0.jpeg) ``` I'll keep testing, but so far, very disappointing…

Why pay more for doctly than an AWS Textract?

I did not try doctly, but AWS Textract does not support in my case Russian, so the output is completely useless