Repository navigation
Closed
15 / 1515 of 15 issues completedLabels
milestone:M5M5 — Intelligent selectionM5 — Intelligent selectiontype:epicMilestone tracking issueMilestone tracking issue
Milestone
Description
Epic / tracking issue for milestone M5 — Intelligent selection · indicative due 2027-01-01 (doc 22 §1)
Goal
RobustnessGP, S-BO, S-LSE, S-PROBE
Exit criteria
- doc 07 §9 benchmark gates pass
- All child tasks below closed
Tasks
- [T-055a] RobustnessGP: SingleTaskGP (Matérn-5/2 ARD) fit/predict/p_fail with fallbacks #110
T-055aRobustnessGP: SingleTaskGP (Matérn-5/2 ARD) fit/predict/p_fail with fallbacks — P0/M - [T-055b] RobustnessGP scale guard: informative subsampling above 1 500 points #111
T-055bRobustnessGP scale guard: informative subsampling above 1 500 points — P1/XS - [T-056a] Acquisitions: qLogNEI wrapper, Expected Violation (EV), straddle #112
T-056aAcquisitions: qLogNEI wrapper, Expected Violation (EV), straddle — P0/S - [T-056b] Diversity penalty over known failure modes + cheap mode cache #113
T-056bDiversity penalty over known failure modes + cheap mode cache — P0/S - [T-056c] Batch proposals: Kriging Believer + acquisition optimiser wrapper (restarts, tabu, dedupe, seeding) #114
T-056cBatch proposals: Kriging Believer + acquisition optimiser wrapper (restarts, tabu, dedupe, seeding) — P0/S - [T-057] S-BO strategy (warm-up + qLogNEI) with branin gate #115
T-057S-BO strategy (warm-up + qLogNEI) with branin gate — P0/S - [T-058] S-LSE strategy (warm-up + straddle level-set estimation) #116
T-058S-LSE strategy (warm-up + straddle level-set estimation) — P1/S - [T-059a] S-PROBE PhaseController: phase selection, weights, ε-explore, adaptive rule (pure) #117
T-059aS-PROBE PhaseController: phase selection, weights, ε-explore, adaptive rule (pure) — P0/S - [T-059b] S-PROBE refine planner: nominal-ward bisection + Gaussian perturbations per new mode #118
T-059bS-PROBE refine planner: nominal-ward bisection + Gaussian perturbations per new mode — P0/S - [T-059c] S-PROBE strategy assembly: ProbeStrategy with rationale, checkpoints and multi_mode_6d gates #119
T-059cS-PROBE strategy assembly: ProbeStrategy with rationale, checkpoints and multi_mode_6d gates — P0/M - [T-060] Strategy params validation and exposure via
/meta(schemas for the wizard) #120T-060Strategy params validation and exposure via/meta(schemas for the wizard) — P1/XS - [T-053] Performance benchmarks (NFR-PERF-001/002) and
benchmark-smokeCI job #121T-053Performance benchmarks (NFR-PERF-001/002) andbenchmark-smokeCI job — P1/S - [T-061] Calibrate benchmark thresholds and freeze S-PROBE hyper-parameters (pre-registration) #122
T-061Calibrate benchmark thresholds and freeze S-PROBE hyper-parameters (pre-registration) — P0/S - [T-064] Weekly
benchmark-full.ymlworkflow (EXP-02 + reduced EXP-01) with regression comment #123T-064Weeklybenchmark-full.ymlworkflow (EXP-02 + reduced EXP-01) with regression comment — P2/S - [T-062] [V1-optional] S-CEM cross-entropy rare-event strategy (plugin) #124
T-062[V1-optional] S-CEM cross-entropy rare-event strategy (plugin) — P3/S
How to work this milestone
- Pick any task whose Blocked by issues are closed. Each task lists the files it owns, so several agents can work in parallel without merge conflicts.
- Follow AGENTS.md, doc 32 (DoR/DoD) and doc 22 §4 (parallelisation).
- Requirements traceability: doc 28.
Metadata
Metadata
Assignees
Labels
milestone:M5M5 — Intelligent selectionM5 — Intelligent selectiontype:epicMilestone tracking issueMilestone tracking issue
Activity
/meta(schemas for the wizard) #120benchmark-smokeCI job #121benchmark-full.ymlworkflow (EXP-02 + reduced EXP-01) with regression comment #123CodewithJha commentedon Nov 22, 2026
Shipped on main — all child tasks closed.