Get the latest tech news
Deterministic Quoting: Making LLMs safer for healthcare
LLMs have the potential to revolutionise our field of healthcare, but the fear and reality of hallucinations prevent adoption in most applications.
LLMs remain imperfect, so it may still choose to quote the wrong part of the source material, but only “real” quotations are displayed on blue - they are deterministically generated. While implementations will vary, DQ fundamentally involves a sending LLM outputs to a separate module that replaces potentially-hallucinated quotations with verbatim copies direct from the source material. Combining the flexibility of LLMs with the reliability of non-AI quote lookups, DQ gives users confidence in the “ground truth” data assumed by the AI.
Or read this on Hacker News
