Campaign session

01KTPR53FRW0A8C0NDN03VYMX4

finished 2026-06-09 17:51:14.936419+00:00 → 2026-06-09 17:55:04.279066+00:00 · 3 runs · supervisor: react-agent

“REPRODUCIBILITY SPOT-CHECK (not a re-resolution). 3 densities x 1 seed on committed-HEAD harness confirms the full session 01KTNHCDZG: (rho,v) = (0.04,1.17),(0.17,1.14),(0.50,0.71), strictly monotonic v-down (Spearman -1.0), matching prior anchors. mean_speed_m_s is populated and speeds physical (0.71-1.17, no degenerate FD) -> the analyser-signature + cyclic-loop scenario fixes are confirmed in COMMITTED code, not just the morning working tree. Criterion 3 (monotonicity) reproduces; criteria 1 (four regimes - geometry-limited) and 2 (in-band fraction) were resolved by the full session and deliberately NOT re-evaluated in this reduced check (marked false = out-of-scope here, not regressed).”

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
Density sweep populates all four Adrian-2024 regimes (>=1 sample each)	no
>=75% of (Voronoi density, Voronoi speed) pairs in the Liddle-2022/Adrian-2024 band	no
(rho,v) varies monotonically with n_agents within each seed: slope negative, per-seed Spearman <= -0.6	yes

Synthesis

c-exps1-fundamental-diagram — session 01KTPR53FR (reproducibility spot-check)

Purpose. Confirm that the EXP-S1 harness fixes made during the full session 01KTNHCDZG (2026-06-09, 06:33–08:25 UTC) — which were flagged there as uncommitted working-tree changes — are now in committed HEAD code and still reproduce the fundamental-diagram result. This is a deliberately reduced 3-point check, not a re-resolution of the brief's criteria.

Verification of committed fixes (code inspection). HEAD contains the cyclic closed-loop journey (jupedsim-runner/entrypoint.py:189–201, replacing the agent-parking central waypoint) and the compute_voronoi_speed PedPy-1.x signature guard (pedpy-analyser/entrypoint.py:215). Working tree clean.

Corpus. 3 runs, n_agents ∈ {4,16,48}, seed 0, base constants (a_seminar / S5 / 180 s / dt 0.05 / transient_frac 0.17). All gate-passed, exit 0, ~9–16 s.

n=4 → ρ=0.042 ped/m², v=1.169 m/s (sim://run/01KTPR5F5EFHTBT202V3B8YX2F)
n=16 → ρ=0.167 ped/m², v=1.139 m/s (sim://run/01KTPR5FFMW4YZZ398NWJXYP1E)
n=48 → ρ=0.500 ped/m², v=0.708 m/s (sim://run/01KTPR5G5ZJ8EHNEGRP9YR5GH3)

Result. Mean Voronoi speed falls strictly monotonically with density (1.17→1.14→0.71 m/s; per-seed Spearman −1.0), and mean_speed_m_s is populated with physical values — no degenerate ~0.04 m/s parking collapse. Densities (0.04/0.17/0.50) match the full session's free-flow + light-congestion anchors. Criterion 3 reproduces exactly. Criteria 1 (four regimes — dense/jam unreachable on this 96 m² geometry) and 2 (in-band fraction) are not re-evaluated here; they were resolved by the full session (2/3) and are unchanged.

Conclusion. The morning session's fixes are durable in committed code; the FD gate is reproducible. No regression. The diary's "analyser emits no mean-speed → blocked" note (2026-06-08) is resolved.

Attached runs

Run	Gate	Purpose	Replay
V3B8YX2F	✓	n4	—
NWJXYP1E	✓	n16	—
P9YR5GH3	✓	n48	—