🧩 2D Differentiable Logic SAT Solver: Product T-Norm Relaxation, Exact Semantic Loss and Gradient Descent
Heat-map view of a differentiable-logic SAT solver: loss over two logits with the descent path, clause lamps and variable probabilities, comparing the fuzzy product t-norm loss with the exact semantic loss, verified by gradient checks, brute-force model counting and a variable-disjoint exactness test.
⚙ Under the hood
Heat-map view of a differentiable-logic SAT solver: loss over two logits with the descent path, clause lamps and variable probabilities, comparing the fuzzy product t-norm loss with the exact semantic loss, verified by gradient checks, brute-force model counting and a variable-disjoint exactness test.
Neurosymbolic AIDifferentiable LogicSemantic LossProduct T-NormWeighted Model Count3-SATGradient DescentLoss Landscape
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install