The AI Committee of the American Society for Indexing is pleased to announce the release of our white paper supplement, “Can the Current Generation of LLMS Produce an Adequate Index?”
Click here to download the white paper in full.
Read the abstract below.
Abstract
This paper supplements our previously-published white paper by investigating the capacity of the current generation of large language models (LLMs) to index a book chapter. The white paper found that an earlier generation of LLMs performed poorly on the three criteria of completeness, navigability, and accuracy. We find that while the current generation does not underindex as severely, it has issues with excessive subheadings and navigability, and continues to fail a standard indexing test for accuracy. LLMs still cannot produce an adequate index.