Show HN: Who's in the weights? – which people 13 language models know
Category: other
Tags: llm-evaluation, data-analysis, machine-learning-research
Score: 6.8/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 5)
This project investigates which people are 'in the weights' of 13 language models by systematically sampling real individuals from Wikidata and comparing model recognition confidence. It's interesting for its rigorous methodology, uncovering artifacts and confounds, and providing honest insights into model behavior and bias.
Target audience: data engineers, ai-ml researchers
Repository: https://whos-in-the-weights.vercel.app/ · Python
View on Hacker News