Developer publishes post-training experiment using ML on demoscene code
1 Oct 31 2:08 PM · 11d ago · 1 article · 2 posts · 3 sources · development 1 of 2
Datavorous published a detailed writeup of training a language model on demoscene and sizecoding projects, specifically 140-character JavaScript demos from Dwitter.net. The project traces visual effects across remixed demos and applies autoregressive language modeling to generate plausible code continuations and explore alternative aesthetic directions.
“Given a partial program, certain characters and expressions are considerably more likely to follow than others. This is pretty much how autoregressive language models work.”
datavorousdatavorous Developer / researcherHellMood Demoscener
The whole story articlespostscomments the bright band is this development · numbered dots are the others · click one to jump
What was reported 1 claim about this development
-
1 outlet I post trained a model on Demoscenes
first by HN Frontpage, 11d ago
All 2 developments of Developer trains language model on demoscene code to… →
Hacker NewsMastodonNewswires