LLMs have to memorize entities. They definitely memorize famous fictional characters. No way it's all of them (is it?). What's the most famous fictional character for which we have trouble finding an associated feature (e.g. neuron, direction, singular value in an MLP, etc.)? Where is the cutoff for memorized knowledge?
If you are inspired by this idea, you can reach out to the authors for collaboration or cite it:
@misc{holtzman-whats-the-most-2026,
author = {Holtzman, Ari},
title = {What's the most famous character that doesn't have an associated feature in an LLM?},
year = {2026},
url = {https://hypogenic.ai/ideahub/idea/IqxOvg2WiLsOG0K4DJ6U}
}Please sign in to comment on this idea.
No comments yet. Be the first to share your thoughts!