/* GDL vs LLM comparison ("What is geometric deep learning?") */ function GdlVsLlm({ accentVar }) { const d = window.LB_DATA.gdlVsLlm; return (
{d.eyebrow}

{d.title}

{d.lead}

{/* Two big headers */}
{/* LLM header */}
ONE APPROACH
Large Language Models
Transformer architectures trained to predict tokens on massive corpora.
{/* GDL header */}
LATENT BIO
Geometric Deep Learning
Networks whose inductive biases respect the geometry data actually lives on.
{/* Rows */} {d.rows.map((r, i) => (
{r.k}
{r.llm}
{r.gdl}
 ))}
Both matter. LLMs are the right tool for language. For biology, we build on its geometry.
); } function LlmGlyph() { // abstract 'token stream' — a row of squares return ( {Array.from({length: 10}).map((_, i) => ( ))} t₁ t₂ t₃ t₄ t₅ t₆ t₇ t₈ t₉ t₁₀ ); } function GdlGlyph({ accentVar }) { // abstract 'manifold graph' — nodes on a curved sheet with edges const nodes = [ [20, 30], [42, 20], [62, 28], [80, 18], [102, 26], [124, 16], [146, 24], [168, 14], [30, 42], [55, 38], [80, 40], [110, 38], [140, 36], [165, 34], ]; const edges = [[0,1],[1,2],[2,3],[3,4],[4,5],[5,6],[6,7],[0,8],[1,9],[2,10],[4,11],[5,12],[7,13],[8,9],[9,10],[10,11],[11,12],[12,13]]; return ( {edges.map(([a,b], i) => ( ))} {nodes.map((n, i) => ( ))} ); } window.GdlVsLlm = GdlVsLlm;