- 137comments
- 154comments
- 16comments
- 249comments
- 148comments
- 4comments
- 25comments
- 27comments
- 15comments
- 364comments
- 3comments
- 10comments
- 26comments
- 94comments
- 71comments
- 128comments
- 169comments
- —discuss
- 26comments
- 26comments
- 33comments
- 205comments
- —discuss
- 6comments
- 33comments
- 34comments
- 32comments
- 209comments
- 94comments
- 89comments
What about citations, and indicating provenance of derived knowledge?
neuml/txtai, neuml/paperai: https://github.com/neuml/paperai
"Wikidata, with 12B facts, can ground LLMs to improve their factuality" (2023) https://news.ycombinator.com/context?id=38309408
Good points, as you reference.
txtai can build citations for single record answers. https://neuml.hashnode.dev/build-rag-pipelines-with-txtai
I have it on my list to figure out how to cite multiple answer/multiple record citations.
Is there different output each time you run this function; does it converge upon true statements predicated by premises?
Is there different output with different system prompts, then?
/? "system prompt" https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...
I haven't noticed that to be the case. Small prompt changes can change the output though as with all LLMs.