Fraser 386614fa68 Finish restructure prompt
I've A/B tested *a lot*, and find doing it this way (sources initially, then truncated conversation, then final question) ends up allowing the user to ask clarifying questions better, with minimal if any drop in citation knowledge and relevance
2023-04-03 20:20:36 -04:00
2023-04-03 20:20:36 -04:00
2023-03-31 20:49:03 -04:00
2023-03-30 03:45:06 -04:00
2023-04-03 20:20:36 -04:00
2023-04-01 01:20:59 -04:00
2023-01-31 04:15:23 -05:00
2023-03-31 20:33:11 -04:00
2023-03-31 20:33:11 -04:00

AlignmentSearch

img

Project Layout

  • src/ stuff around processing our dataset and constructing embeddings
  • api/ a flask app serving as our backend
  • web/ a NextJS app for the frontend
S
Description
Conversational chatbot to answer questions about AI Safety & Alignment based on information retrieved from the Alignment Research Dataset
Readme MIT
53 MiB
Languages
Python 52.6%
TypeScript 43.1%
JavaScript 2.1%
CSS 1.3%
Shell 0.6%
Other 0.3%