// Drive quantitative analysis, factor diagnostics, and reporting for ATFT-GAT-FAN outputs.
| name | atft-research |
| description | Drive quantitative analysis, factor diagnostics, and reporting for ATFT-GAT-FAN outputs. |
| proactive | true |
ls -lt runs | head.python scripts/research/summarize_run.py --run runs/<timestamp>.make research-baseline RUN=runs/<timestamp> — compares to curated benchmark.make research-plus RUN=runs/<timestamp> — full bundle (feature importance, turnover, drawdowns).python scripts/research/plot_metrics.py --run runs/<timestamp> --horizons 1 5 10 20.python scripts/research/graph_analytics.py --dataset output/ml_dataset_latest_full.parquet.reports/<timestamp>/.docs/research/weekly_digest.md.python scripts/research/factor_drift.py --window 60 --features top50.python scripts/research/check_leakage.py --dataset output/ml_dataset_latest_full.parquet.python scripts/research/regime_detector.py --regimes 4 --method gaussian_hmm.python scripts/research/evaluate_by_regime.py --run runs/<timestamp> --regime-file output/regimes/latest.parquet.python scripts/research/limit_checker.py --run runs/<timestamp> — verifies VAR, exposure, and shorting constraints.pytest tests/research/test_safety_constraints.py -k exposure if guard fails.make research-report FACTORS=returns_5d,ret_1d_vs_sec HORIZONS=1,5,10,20.python scripts/research/notebooks/render.py docs/notebooks/performance_atlas.ipynb.python tools/chart_creator.py --input reports/<timestamp>/summary.json --output outputs/figures/.output/ml_dataset_latest_full.parquetruns/<timestamp>/predictions.parquetdataset_features_detail.jsondata/benchmarks/nikkei225.parquetdocs/research/archive/<YYYY-MM-DD>_run_<timestamp>.md../tools/codex.sh "Generate new factor hypothesis from latest run" to synthesize research leads using Codex search + reasoning stack.codex exec --model gpt-5-codex "Summarize regime analysis findings in docs/research/weekly_digest.md" for automated reporting drafts.