LLMs for structured input validation

4 points by kevalshah90 1y ago ↗ HN

There are several examples of structured outputs from LLMs; however, I'd like to use an language model to see if a given document / pdf adheres to a set template. Are there any good examples of this input validation use-case?

If the input doc / text doesn't adhere to the specific template, I'd like LLMs to point our the inconsistencies.

5 comments

[ 0.88 ms ] story [ 23.5 ms ] thread

ailef 1y ago ↗

What would be the advantage of using an LLM here? As opposed to parsing/validating the input using any other technique (of it's a well defined format, where you have e.g. a grammar).

vismit2000 1y ago ↗

Agree with this. Just use Pydantic.

[dead] MarcAntonyX 1y ago ↗

[dead]

pizza 1y ago ↗

Depends on how fuzzy your definition of template is. If it's really structured, a simpler algorithm might work better. Otherwise maybe try https://github.com/Unstructured-IO/unstructured

luke-stanley 1y ago ↗

You might want to try Instructor with llm_validator, where you can have natural language clauses to check against in a loop: https://python.useinstructor.com/blog/2023/10/23/good-llm-va... With Pydantic too of course!