31–40 of 51 posts
Earlier quoted context omitted.
To be fair, they didn't include themselves at all in the graph.
They did. It’s in the #1 spot Update: looks like the removed themselves from the graph since I saw it earlier today!
The beauty of version control: https://github.com/getomni-ai/benchmark/commit/0544e2a439423...
Earlier quoted context omitted.
So not all of them are local and open source? Ugh.
I don't see why you couldn't run any of those locally if you buy the right hardware?
If they (all of the mentioned ones) are open source and can be ran locally, then most likely, yes.
From what I remember, they are all local and open source, so the answer is yes, if I am correct.
How does one configure an LLM interface using this to process multiple files with a single prompt?
How does one configure an LLM interface using this to process multiple files with a single prompt?
(I would recommend the latter)
I suppose none of these models can output bounding box coordinates for extracted text? That seems to be a big advantage of traditional OCR over LLMs. For applications I'm interested in, until we can get to 95+% accuracy, it will require human double-checking / corrections, which seems unfeasible w/o bounding boxes to quickly check for errors.
I suppose none of these models can output bounding box coordinates for extracted text? That seems to be a big advantage of traditional OCR over LLMs. For applications I'm interested in, until we can get to 95+% accuracy, it will require human double-checking / corrections, which seems unfeasible w/o bounding boxes to quickly check for errors.
qwen2.5-vl-72b-instruct seems perfectly happy outputting bounding boxes in my testing. There's also a paper https://arxiv.org/pdf/2409.12191 where they explicitly say some of their training included bounding boxes and coordinates.
I suppose none of these models can output bounding box coordinates for extracted text? That seems to be a big advantage of traditional OCR over LLMs. For applications I'm interested in, until we can get to 95+% accuracy, it will require human double-checking / corrections, which seems unfeasible w/o bounding boxes to quickly check for errors.
Actually qwen 2.5 is trained to provide bounding boxes
https://github.com/QwenLM/Qwen2.5-VL/blob/main/cookbooks/ocr...