[2605.26492] Elias in the Lighthouse, Again? Diagnosing Low Diversity in LLM Stories
Researchers Sil Hamilton and David Mimno found that stories generated by large language models show low diversity. They discovered that 11 words appear in 88.3% of the sampled 20,000 stories, revealing a reliance on limited vocabulary influenced by preference data. This pattern raises concerns about how small datasets can disproportionately shape the narratives produced by these models.
This is worth holding only if the practical relevance is clear from the source.
This record is extracted from a published AI Today issue and tied to the original source URL. Treat the source as the record of evidence for the summary.