NORTH-STAR · 북극성

Follow a hand-written ruler with a learned recursion — from few data.

Train one system out of two objects: ① a ruler Ec that defines what "correct" means, and ② GRAM, a weight-shared recursion that builds the answer over K=10 steps. Goal — teach ② to obey ① from few examples, no augmentation, and generalize to held-out and OOD.

Ruler · Ec

the verifier

Hand-written row/col/box conflict count. Exactly 0 only at a valid solution.

×

GRAM

the recursion

Weight-shared, K=10 steps. Builds the answer; learns to obey ① from few data.

The law. Energy is consumed as a discrete-state OBSERVATION / verifier — never descended as a continuous −∇E gradient field. Energy-as-gradient died 8× (non-navigable, two-basin, saddle); energy-as-observation lives.

0.75
weakOOD exact · ①×② verified search (record)
0 → 0.31
Held-out exact · greedy 0.07 → model-guided search
Search-budget efficiency from ② branch ordering
0 → 0.09
targetOOD exact · backtracking wall opened
CA-2 multi-path distillation running collect verified search traces (6 shards) → distill into the operator · trace-p 0.5 live

Honest status. Best held-out exact 0.090 is still below the propagation ceiling 0.150 — the learned operator is a soft statistical corrector, not yet a sound propagator. targetOOD exact = 0 (needs guess + backtrack). This list extends as dates grow.