A horizontal banner image depicting a snowy mountain.

marius.vision

Software Ghostbox Tanto AI Museum B/X Toolkit
Media HeMakesMePlay A Dungeon About Blue Sky Github

AI Museum

The problem of evaluating LLMs has never been a simple one. Direct metrics like perplexity or KL-divergence are objective and comparable, but fail to capture a model's vibe. Benchmarking approaches based on task-effectiveness can be gamed and sometimes clash with people's real world experience of a model. Indeed, comprehensively capturing the intelligence of an artificial intelligence remains entirely elusive.

New Perspectives

The AI museum is an attempt to approach this issue within a new paradigm. Where before we sought to evaluate, the AI museum invites the visitors gaze to appreciate.

In the AI museum you will find several exhibits that are all generated by various transformer models.

A refreshing change of Pace

Choose one of the exhibits above to begin your journey through the fascinating world of AI generated content.


© 2026 marius.vision All rights reserved. Powered by get-off-my-lawn.Admin.RSS Icon