[cschanhniem]

Change

Add active-learning strategy comparison report (openamp-foundry bench strategy-compare) that compares 5 selection strategies (exploitation, exploration, diversity, combined, random) on the same synthetic pool with identical hidden active candidates.

Why

Prevents one-selector bias by making strategy performance transparent. Until now, only the combined production selector was benchmarked against random. This report lets reviewers see whether simpler strategies (pure exploitation, pure exploration, pure diversity) outperform or underperform the production selector on the same recovery task.

What

• src/openamp_foundry/active_learning/strategy_comparison.py — run_strategy_comparison() generates synthetic pool, runs 5 strategies, produces structured comparison report with per-strategy metrics, ranking by recall, and production selector vs pure-strategy comparisons.
• schemas/active_learning_strategy_comparison.schema.json — JSON Schema Draft 2020-12.
• CLI: openamp-foundry bench strategy-compare with flags for pool size, hidden actives, batch size, rounds, and JSON/Markdown output.
• make bench-strategy-compare Makefile target.
• 18 tests covering structure, ranking, output formats, schema, CLI, and edge cases.
• Docs: METRICS_CURRENT.md, ROADMAP.md, test count baseline (2955).

Evidence

• 18 new tests all pass
• Test count regression gate updated to 2955
• make agent-check passes
• CLI produces valid JSON + Markdown output

Safety impact

None. This is a computational comparison report using synthetic data. No biological claims, no release changes, no policy changes.

Proof level

N/A — no scientific claims. Informational comparison of algorithm variants.

Limitations

• All data is synthetic. Results reflect code-path integrity, not biological performance.
• Strategies run with safety/selectivity gates disabled to isolate strategy effects.
• Random baseline is 20-trial Monte Carlo average; deterministic strategies are single-run.
• The production selector optimizes for multiple objectives beyond recall.

Human review needed

No — informational infrastructure only.