Working Papers

Formal tracks for the OptFin research program

This page gives the public research shape of the lab: theory, benchmarks, exactness, biological applications, and challenge-driven engineering. It is the bridge between notes and formal papers.

Theory

The OptFin structural hypothesis

A knowledge-bearing AI can repeatedly collapse decisive subgraphs, solve or tightly bound them, and lift those results back into the original problem until lower and upper bounds converge.

Benchmarks

ROADEF and historical industrial regression

Historical challenge families remain the proving ground for a serious optimization engine: assignment, machine reassignment, packing, scheduling, routing, and adaptive segment routing.

Biodesign

Exact subgraph refinement for molecular specificity

Generative systems propose candidates; local exactness verifies and sharpens the specificity frontier in critical interfaces.

Proof

Feasibility, gap, and objective as publication requirements

Every report should preserve the instance, feasibility status, objective value, lower bound, upper bound, gap, runtime, and stopping reason in one reproducible bundle.