So we've been digging into ways to reduce hallucinations and make LLM's output more robust and deterministic. We're seeing some crazy good early results in our test runs 🤞
If you're running into any issues with prompt/output quality, just DM and we can chat!
Jeng Yang Chia
08/21/2023, 11:40 PM
We've been feeling v. frustrated on this bit. Been building bots in the travel space. Thinking of adding knowledge graphs to our bots. How are you guys approaching it?
08/21/2023, 11:42 PM
Hey Jeng, thanks for getting back! DM'ing you
08/28/2023, 2:59 PM
I’ve been considering going back and refactoring all my openai prompts to use functions so that I can more rigidly enforce structure, namely preserving citations from search results.
Curious how the knowledge graphs are working for you @Jeng Yang Chia?