in-context-learning

1 article
sort: new top best
clear filter
0 2/10

This paper proposes using Neural Cellular Automata (NCA)—synthetic data generated from learned transition rules on grids—as pre-training data for language models, achieving 6% perplexity gains and 1.6× faster convergence than natural language pre-training at equivalent scale. The key insight is that NCA sequences force models to develop in-context rule inference capabilities purely from structural patterns without semantic shortcuts, resulting in more transferable representations to downstream language tasks.

Neural Cellular Automata (NCA) OpenWebText OpenWebMath CodeParrot C4 GSM8K HumanEval BigBench-Lite Conway's Game of Life
hanseungwook.github.io · Anon84 · 17 hours ago · details · hn