Search papers, labs, and topics across Lattice.
This study investigates whether syntactic processing can emerge from local interactions in a neural cellular automaton (NCA) with 18,658 parameters, trained solely on a 1-bit boundary signal to solve the membership problem of an arithmetic-expression grammar. The NCA self-organizes into a representation termed Proto-CKY, which exhibits expressive power beyond regular languages and demonstrates structural generalization while maintaining a quantitative alignment with grammatical structure (Pearson \( r \approx 0.71 \)). Notably, Proto-CKY emerges consistently across four context-free grammars and can regenerate after perturbations, highlighting its robustness and functional alignment with the CKY algorithm despite formal differences.
Syntactic processing can spontaneously emerge from local interactions, revealing a novel representation that aligns with grammatical structures and defies conventional training paradigms.
Can syntactic processing emerge spontaneously from purely local interaction? We present a concrete instance on a minimal system: an 18,658-parameter two-dimensional neural cellular automaton (NCA), supervised by nothing more than a 1-bit boundary signal, is trained on the membership problem of an arithmetic-expression grammar. After training, its internal $L \times L$ grid spontaneously self-organizes into an ordered, spatially extended representation that we name Proto-CKY. This representation satisfies three operational criteria for syntactic processing: expressive power beyond the regular languages, structural generalization beyond the training distribution, and an internal organization quantitatively aligned with grammatical structure (Pearson $r \approx 0.71$). It emerges independently on four context-free grammars and regenerates spontaneously after perturbation. Proto-CKY is functionally aligned with the CKY algorithm but formally distinct from it: it is a physical prototype, a concrete instantiation of a mathematical ideal on a physical substrate, and the systematic distance between the two carries information about the substrate itself.
