Paper Title (Year) — Short Tagline

1. Problem & Tasks

• What is being predicted/estimated?
• Benchmarks/tasks, evaluation settings (classification, regression, survival, etc.).
• How this connects to genetics × MRI integration (if applicable).

2. Datasets

• Cohorts, sample sizes, populations, modalities.
• Preprocessing specifics (QC, parcellations, sequencing protocols, etc.).
• Splits: train/val/test definitions, CV type, site-aware grouping, seeds.

3. Model / Method Details

3.1 Architecture / Method

• Backbone, context length, tokenization, RC-equivariance, hub tokens, etc.
• Input/output formats, objectives, losses, notable tricks.

3.2 Confound Handling & Evaluation Discipline

• Residualization, stratification, harmonization, missingness handling.
• Metrics (AUROC/AUPRC/Brier/calibration), statistical tests (DeLong, bootstrap, permutations, FDR).

4. Results & Tables

• Key numbers vs baselines (include effect sizes and uncertainty if reported).
• Ablation results or thresholds that matter for reuse.

5. Limitations & Cautions

• Author-listed limitations plus any reuse caveats you notice.
• Cohort bias, preprocessing quirks, compute constraints, licensing limits.

---

6. Hooks into Neuro-Omics KB

Relevant KB assets

• kb/paper_cards/<slug>_YYYY.yaml
• kb/model_cards/... (if applicable)
• kb/datasets/... (if applicable)
• docs/generated/kb_curated/integration_cards/<related>.md

Configs / recipes informed

• configs/experiments/...
• docs/integration/analysis_recipes/...
• docs/integration/integration_strategy.md

Concrete guidance for our project

• Fusion pattern / modality sequencing choices this paper justifies.
• Specific covariates, statistical tests, or dataset fields we adopted.
• Any LOGO/PRS/GWAS/CCA protocol parameters we ported over.

Copy this file to docs/generated/kb_curated/papers-md/<slug>.md and fill in the sections. Keep Layer 2 MDs canonical, so future tooling (RAG, dashboards) can scan them systematically.