← Back to context

Comment by zacmps

17 days ago

I've found the highest accuracy solution is to OCR with one of the dedicated models then feed that text and the original image into an LLM with a prompt like:

"Correct errors in this OCR transcription".

7 comments

zacmps

Reply
therein  17 days ago

How does it behave if the body of text is offensive or what if it is talking about a recipe to purify UF-6 gas at home? Will it stop doing what it is doing and enter lecturing mode?

I am asking not to be cynical but because of my limited experience with using LLMs for any task that may operate on offensive or unknown input seems to get triggered by all sorts of unpredictable moral judgements and dragged into generating not the output I wanted, at all.

If I am asking this black box to give me a JSON output containing keywords for a certain text, if it happens to be offensive, it refuses to do that.

How does one tackle with that?

  • devjab  17 days ago

    We use the Azure models and there isn't an issue with safety filters as such for enterprise customers. The one time we had an issue microsoft changed the safety measures. Of course the safety measures we might meet are the sort of engineering which could be interpreted as weapons manufacturing, and not "political" as such. Basically the safety guard rails seem to be added on top of all these models, which means they can also be removed without impacting the model. I could be wrong on that, but it seems that way.

  • xnx  17 days ago

    There are many settings for changing the safety level in Gemini API calls: https://ai.google.dev/gemini-api/docs/safety-settings

  • zacmps  17 days ago

    It's not something I've needed to deal with personally.

    We have run into added content filters in Azure OpenAI on a different application, but we just put in a request to tune them down for us.