Contextless context is not useful.
- Contextless context is not useful.
- Context that is incorrect is worse than nothing.[hr]
- Randomizing things the LLM will pay attention to, overriding its intuition on the right answer.
- If you give it good context it nudges in a useful way.
- If you give it bad context it overwhelms the model and just returns the context.
- It overshadows the answer.
- An example in the wild:
- ""come up with some business ideas for untapped markets in my area"
- >Based on your love of astronomy and that one time 8 months ago you mentioned you like watching basketball, so how about a youth basketball coach and stargazing guide!"[hs]