sequential-edge-inverse-entropy-voting

Replace parallel self-consistency with sequential reasoning where chains iteratively build on previous attempts, weighted by inverse entropy to prioritize confident solutions, achieving 46.7 pp accuracy gains over parallel approaches.

Independent GitHub snapshot · Automated classification · Original documentation governs