docs/p1_transfer/p1_reward_transfer.md

P1 reward transfer for fusion-design-lab

Core conclusion

Use scadena as the main repair prior and CreativeEngineer as the top-score exploitation prior.

Reward schedule

1. Infeasible phase: reduce official feasibility violation first.
2. Within infeasible candidates, prioritize average triangularity repair first.
3. After triangularity starts to clear, prefer aspect-ratio and edge-iota cleanup.
4. Crossing the official feasibility threshold should receive a large bonus.
5. Once feasible, optimize the official P1 score, which in practice tracks lower max elongation.

Practical priors

• scadena_seed.json: best repair anchor. Raw near-P1 HF seeds crossed official feasibility by moving toward this family.
• creative_seed.json and creative_best.json: best endpoint family. Use these for exploitation after feasibility is stable.
• samet_seed.json: exact-feasible distinct family. Useful for diversity and non-scadena exploration.
• egodos_seed.json: best near-feasible non-scadena source.
• egodos_sparse_rgroup_best.json: proof that a sparse low-order r_cos move can create a new feasible non-scadena design.

Operators that worked

• Coordinated move toward the scadena manifold for repairing raw near-P1 seeds.
• Sparse low-order r_cos move from egodos toward Samet.
• Small local Samet continuation on low-order modes.

Operators that did not work well

• Whole-boundary interpolation between distant families.
• Random full-space coefficient noise.
• Sparse top-k replacement toward scadena without coordinated movement.

Suggested usage in fusion-design-lab

• During repair-heavy exploration, add reward telemetry for:
  • official feasibility
  • average triangularity
  • aspect ratio
  • edge rotational transform over field periods
  • max elongation
• Bias mutation proposals toward:
  • scadena direction for feasibility repair
  • CreativeEngineer neighborhood for high-score exploitation
  • Samet and egodos seeds for diversity maintenance