Campaign session

01KX3Y9PYNEKZ4PZKDC0YFVS84

finished 2026-07-09 17:20:08.661219+00:00 → 2026-07-09 17:24:25.746444+00:00 · 1 run · supervisor: react-agent

“Preflight PASSED and launch-recipe proven; powered 1650-run corpus deferred to a dedicated long-running CI dispatch (insufficient LLM budget this instance to supervise it to completion). No science criterion resolved this session.”

Archive snapshot, as of 1 day ago — the run corpus is rebuilt once a day, so this page is not a live reading. The fleet panel is the live one; it refreshes every 30 s.

Success criteria

Criterion	Resolved
Primary: bootstrap-over-floors 95% CI on LOFO inflation ratio (held-out MAE / within-floor leave-one-occupancy-out MAE), lower bound > 1.5x.	no
Variance split: nested decomposition reports sigma2_floor and sigma2_seed; geometry-dominates iff between-floor SD of per-floor CV(N) exceeds within-floor between-seed SD.	no
Comparator honesty: LOFO denominator is the held-out within-floor baseline (leave-one-occupancy-out), never in-sample training residual.	no

Synthesis

Session 01KX3Y9PYNEKZ4PZKDC0YFVS84 — preflight-validated, powered corpus deferred

Outcome: clean partial (BLOCKED-on-budget, NOT blocked-on-infra). This session executed the mandatory Step-2.5 preflight for the powered Phase-2 scale-out and stopped short of the 1650-run corpus on an honest LLM-budget limit. No science criterion was tested; all three remain open.

What was established (durable)

PostGIS reachable — gis_stats returned 113 floors / 116 experiments. The sibling session's standing blocker (connection-refused on :5435, 2026-07-06) is cleared on this runner.
Execution chain live — a probe exp-csi-static run (floor resplan-1041-floor-0, n_agents=0, 8 placements, seed 0, where=local) gate-passed: exit 0, 5.7 s, 7 artefacts. gis export-floor scene-staging works end-to-end.
Launch recipe proven — a dry_run sim_sweep_launch over one replication expanded to the correct 15 cells (5 occupancy × 3 seeds × 1 freq), with experiment↔floor paired as single-value axes and nested radio.carrier_freq_hz/body.* dotted axes resolved. This is the §2z path whose misuse zero-run-partialed the prior CI session 01KWZC9E…; it is now verified correct.

Criteria status

C1 (powered inflation-ratio CI): unresolved — requires the ≥40-floor corpus (n₀≈40 at σ≈4.82, d=1.5). Not run.
C2 (variance split σ²_floor/σ²_seed): unresolved this session; the pilot (session 01KWYBEJQK, 10 floors) is the only extant estimate.
C3 (comparator honesty): the honest held-out leave-one-occupancy-out LOFO denominator is present in the lofo_cv reduction primitive + csi_cross_geometry_scaleout.py contract, but was not exercised on a corpus.

Recommendation

Dispatch the powered run as a dedicated long-running job (/campaign-systematic, budget_factor=1.0, full 150 k-token / 24 CPU-h grant) — the brief's own prescription. Corpus: 110 replications × 15 cells = 1650 runs, one sim_sweep_launch per paired replication (recipe validated above), reduced via lofo_cv + nested variance decomposition, then the two mandatory figures (csi_cross_geometry small-multiples + CI panel; seed_trace). No infra work is needed first — the chain is confirmed green.

Attached runs

Run	Gate	Purpose	Replay
HZBZ1MGG	✓	—	—