Re:Ask HN: Strategies to Reduce AI Hallucinations?
by jhanschoo on Monday July 29, 2024 @05:07AM
(#41099014)
Some basic theoretical intuition behind this:
the compute graph for an LLM given context before it has to emit a token is traditionally bounded. By asking an LLM to generate intermediate answers, and have that visible as context, you are essentially allowing the LLM to use more compute before it gives you an answer. My guess is that SOTA models are far more flexible in that regard.
the compute graph for an LLM given context before it has to emit a token is traditionally bounded. By asking an LLM to generate intermediate answers, and have that visible as context, you are essentially allowing the LLM to use more compute before it gives you an answer. My guess is that SOTA models are far more flexible in that regard.