# Evidence gallery

27 screenshots. Every image is either a `sumo-gui` screenshot of the running simulation, taken over TraCI during an audit trial, or a screenshot of the audit dashboard rendering recorded result files. None is drawn, edited or composited. Each image's sha256 was re-verified on disk when this page was generated (all match).

Reading the simulator images: the thin red, yellow and green lines at the stop lines are the signal states SUMO is showing; the cyan bars just past junctions are the lane-area detectors the harness uses to measure spillback; the ambulance is white with a cross icon, the fire engine red, the police car blue, ordinary cars yellow. A simulator image shows the state at `sim time`, one step after the capture call.

## E01: city36, all 36 signalised junctions (C00-C55), VECTOR controlling every one, t=301 s

![E01](../../../artifacts/final-adversarial-audit/evidence/E01.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `A/vector/95000` (scenario A, arm vector, seed 95000); sim time 301.0 s; captured 2026-10-25T18:17:19.175518+00:00.
* **Claim it supports:** the evaluated network is the 36-junction grid, under live control
* **Limitation:** an image shows layout and traffic; junction count is proven by topology_report.json
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/A/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/A_vector_95000.jsonl.gz`
* **sha256:** `5707b9171dbdc31218683eb37fde0122f39c1e9a4a919cde14f2c0de45bfdb13`
* **State recorded at capture:** halting on approaches 56; C33: YELLOW EW (last decision: VEHICLE_DEMAND (1 vehicles in lane_1; gap_out)); C32: GREEN EW (last decision: NO_DEMAND_REST_IN_GREEN); C23: GREEN NS (last decision: NO_DEMAND_REST_IN_GREEN)

## E02: Topology validation of city36: 36 signal-controlled intersections, gate A

![E02](../../../artifacts/final-adversarial-audit/evidence/E02.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `—` (scenario —, arm —, seed —); captured 2026-10-25T18:33:20.839Z.
* **Claim it supports:** the executed network has 36 connected signal-controlled intersections, validated from the SUMO network files
* **Limitation:** a synthetic 6x6 grid, not a real city; validation is structural (junctions, links, programmes), not behavioural
* **Linked data:** `artifacts/final-adversarial-audit/topology/topology_report.json`
* **sha256:** `d3f8ab9ab508884ab7eeb12a94a6d2abf607d8ad6d1e3be73c610d1dcb2d5bbc`

## E03: the same corridor, same seed and surge, under SUMO's native actuated baseline

![E03](../../../artifacts/final-adversarial-audit/evidence/E03.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `A/native_actuated/95000` (scenario A, arm native_actuated, seed 95000); sim time 1201.0 s; captured 2026-10-25T18:04:42.728208+00:00.
* **Claim it supports:** the baseline faces identical demand at the same instant
* **Limitation:** one instant of one seed; the comparison is in the metrics, not in two pictures
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/A/native_actuated/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/A_native_actuated_95000.jsonl.gz`
* **sha256:** `e38b744a138a1b536165ad2bb2ba02a86d397d09fa88e31e4facdc204e0a7776`
* **State recorded at capture:** halting on approaches 220; C31: ALL_RED; C21: ALL_RED; C30: GREEN EW

## E04: Scenario G, junction C00 (congested corner), t 1688-1726 s: true queue, what the sensor model showed VECTOR, the QueueObserver belief and the decision

![E04](../../../artifacts/final-adversarial-audit/evidence/E04.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `G/vector/95000` (scenario G, arm vector, seed 95000); sim time 1688 s; captured 2026-10-25T18:33:21.378Z.
* **Claim it supports:** VECTOR decided on corrupted observations (noise, dropouts, phantom spikes); the observer clamps impossible jumps
* **Limitation:** the sensor model is the audit's own simple model (Gaussian noise, Bernoulli dropout, spikes), not a real detector's failure physics
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/G_vector_95000.jsonl.gz`
* **sha256:** `418fe6a5392029f97eeb21fb459ba1da81ec2d494fd924b4ed87c7d50d2a78fc`

## E05: Scenario R0, junction C21: shadow forecast of total queue at t+8 s against what happened

![E05](../../../artifacts/final-adversarial-audit/evidence/E05.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `R0/vector/95000` (scenario R0, arm vector, seed 95000); captured 2026-10-25T18:33:21.991Z.
* **Claim it supports:** the shadow forecasters run in the live loop and are scored against the realised queue
* **Limitation:** shadow only: no forecast reaches a signal decision; held-out MAE is worse than persistence in 15 of 16 scenarios
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/R0_vector_95000.jsonl.gz`
* **sha256:** `ecf7915ba59981691bf3171958be62db07ae04926227b8e1015a4d0b61365d1e`

## E06: Scenario H, junction C21: shadow forecast across the concept shift at 900 s

![E06](../../../artifacts/final-adversarial-audit/evidence/E06.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `H/vector/95000` (scenario H, arm vector, seed 95000); captured 2026-10-25T18:33:22.562Z.
* **Claim it supports:** prediction error rises after the demand shift; nothing adapts online
* **Limitation:** one junction and one seed are shown; the aggregate is in prediction.json
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/H_vector_95000.jsonl.gz`, `artifacts/final-adversarial-audit/results/heldout/prediction.json`
* **sha256:** `0e39288d3d853fea6e608fe1324806f7599a54dc43a156b09fcbe21cd4f5c7d2`

## E07: row-1 corridor at 1201 s, during the x4 surge from W1 and E1 (600-1500 s): the queue is concentrated at the western entry C01 (52 halting on EW); the downstream corridor junctions hold 0-3

![E07](../../../artifacts/final-adversarial-audit/evidence/E07.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `A/vector/95000` (scenario A, arm vector, seed 95000); sim time 1201.0 s; captured 2026-10-25T18:17:51.539483+00:00.
* **Claim it supports:** the surge congests the corridor entry, and VECTOR's input there is the halting count
* **Limitation:** VECTOR has no congestion detector; 'detection' here is the halting count it is shown
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/A/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/A_vector_95000.jsonl.gz`
* **sha256:** `d690b5bf344b0d43277d13baf395189d83d83b08c1d062d6d2964a8bc875eecf`
* **State recorded at capture:** halting on approaches 297; C31: GREEN EW (last decision: VEHICLE_DEMAND (2 vehicles in lane_2; discharging 2)); C21: GREEN EW (last decision: VEHICLE_DEMAND (3 vehicles in lane_2)); C30: GREEN NS (last decision: VEHICLE_DEMAND (6 vehicles in lane_1))

## E08: C21 at its first commanded change after 1200 s: after clearance VECTOR gives EW green to 3 halting vehicles (VEHICLE_DEMAND); C11, left, is in all-red after ending EW green on gap-out

![E08](../../../artifacts/final-adversarial-audit/evidence/E08.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `A/vector/95000` (scenario A, arm vector, seed 95000); sim time 1201.0 s; captured 2026-10-25T18:17:51.543470+00:00.
* **Claim it supports:** a VECTOR decision was executed by SUMO (commanded = reported state, checked each step)
* **Limitation:** the reason is in the decision record beside the image, not in the image
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/A/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/A_vector_95000.jsonl.gz`
* **sha256:** `d690b5bf344b0d43277d13baf395189d83d83b08c1d062d6d2964a8bc875eecf`
* **State recorded at capture:** halting on approaches 297; C21: GREEN EW (last decision: VEHICLE_DEMAND (3 vehicles in lane_2)); C11: ALL_RED (last decision: VEHICLE_DEMAND (1 vehicles in lane_1; gap_out)); C20: GREEN NS (last decision: VEHICLE_DEMAND (1 vehicles in lane_1))

## E09: row-1 corridor, same instant: six junctions, six independent decisions

![E09](../../../artifacts/final-adversarial-audit/evidence/E09.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `A/vector/95000` (scenario A, arm vector, seed 95000); sim time 1211.0 s; captured 2026-10-25T18:17:52.635789+00:00.
* **Claim it supports:** each corridor junction responds to its own queues
* **Limitation:** there is NO coordination mechanism (no offsets, no shared state); this shows its absence, not coordination
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/A/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/A_vector_95000.jsonl.gz`
* **sha256:** `ee081fd837e930f83d5a418e1a934377f40078f0fe52c5d558596f80803d658c`
* **State recorded at capture:** halting on approaches 321; C31: YELLOW EW (last decision: VEHICLE_DEMAND (6 vehicles in lane_1; gap_out)); C21: GREEN EW (last decision: NO_DEMAND_REST_IN_GREEN); C30: GREEN NS (last decision: VEHICLE_DEMAND (3 vehicles in lane_1; discharging 3))

## E10: ambulance amb1 (white, eastbound on row 1) approaching C21; the decision record beside it shows whether VECTOR gave its approach priority

![E10](../../../artifacts/final-adversarial-audit/evidence/E10.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `C/vector/95000` (scenario C, arm vector, seed 95000); sim time 947.0 s; captured 2026-10-25T18:05:29.913635+00:00.
* **Claim it supports:** VECTOR's emergency rule acts on an ambulance it is shown
* **Limitation:** detection is ideal (simulator ground truth, confidence 0.95); vehicles do not yield
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/C/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/C_vector_95000.jsonl.gz`
* **sha256:** `0dbd4c40bcafc2aa8f49931c10bb571d5ed807c9ed11b77d4ce38763aaf845de`
* **State recorded at capture:** halting on approaches 130; C21: GREEN EW (last decision: EMERGENCY_OVERRIDE (0.95)); C11: GREEN NS (last decision: VEHICLE_DEMAND (6 vehicles in lane_1)); C20: GREEN NS (last decision: VEHICLE_DEMAND (5 vehicles in lane_1)); emergency vehicles {"amb1": {"x": 542.6, "y": 375.2, "speed": 13.48, "lane": "C11_C21_0"}}

## E11: fire engine fire1 (red, northbound on column 2) after the C21 conflict, approaching C22

![E11](../../../artifacts/final-adversarial-audit/evidence/E11.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `D/vector/95000` (scenario D, arm vector, seed 95000); sim time 960.0 s; captured 2026-10-25T18:06:40.863370+00:00.
* **Claim it supports:** the fire engine is handled by the same emergency rule
* **Limitation:** one vehicle at one instant; its full journey is in the EV table
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/D/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/D_vector_95000.jsonl.gz`
* **sha256:** `d19491dfb6a859d43f97ebc698adda30e813ec0d7a3666979daa7c30ef9960c5`
* **State recorded at capture:** halting on approaches 136; C22: ALL_RED (last decision: EMERGENCY_OVERRIDE (0.95)); C12: GREEN EW (last decision: NO_DEMAND_REST_IN_GREEN); C23: GREEN EW (last decision: VEHICLE_DEMAND (1 vehicles in lane_2)); emergency vehicles {"amb1": {"x": 715.6, "y": 378.4, "speed": 13.01, "lane": "C21_C31_1"}, "fire1": {"x": 601.6, "y": 551.5, "speed": 10.91, "lane": "C21_C22_0"}}

## E12: C21 as the ambulance (EW) and the fire engine (NS) converge on conflicting approaches

![E12](../../../artifacts/final-adversarial-audit/evidence/E12.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `D/vector/95000` (scenario D, arm vector, seed 95000); sim time 938.0 s; captured 2026-10-25T18:06:39.989491+00:00.
* **Claim it supports:** competing requests at one junction: the record shows which approach was served
* **Limitation:** VECTOR's input admits ONE emergency per junction; the harness reports the most recently dispatched one, so the other gets no priority
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/D/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/D_vector_95000.jsonl.gz`
* **sha256:** `0b3131306852c7fc8bcc83687f953e9ea4ce0336177b198ecde240d226a28443`
* **State recorded at capture:** halting on approaches 134; C21: GREEN NS (last decision: EMERGENCY_OVERRIDE (0.95)); C11: YELLOW EW (last decision: VEHICLE_DEMAND (5 vehicles in lane_1; gap_out)); C20: GREEN EW (last decision: VEHICLE_DEMAND (16 vehicles in lane_2)); emergency vehicles {"amb1": {"x": 423.2, "y": 375.2, "speed": 12.81, "lane": ":C11_13_0"}, "fire1": {"x": 601.6, "y": 312.9, "speed": 10.73, "lane": "C20_C21_0"}}

## E13: capacity reduction on C12->C13 (column-1 arterial northbound), 1001 s: a vehicle stalled at C13's stop line in lane 0 since 700 s, speed limit 30% on the link; traffic passes in lane 1

![E13](../../../artifacts/final-adversarial-audit/evidence/E13.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `F/vector/95000` (scenario F, arm vector, seed 95000); sim time 1001.0 s; captured 2026-10-25T18:44:35.959196+00:00.
* **Claim it supports:** the incident removes one of two lanes at C13's northbound stop line
* **Limitation:** VECTOR is not told of the incident; it sees only queues (the stalled vehicle is excluded from the count it is shown)
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/F/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/F_vector_95000.jsonl.gz`
* **sha256:** `cadee3ee103686b52a32496708d6d7fc17572f244ee453e06157629b7b440b43`
* **State recorded at capture:** halting on approaches 124; C13: GREEN NS (last decision: VEHICLE_DEMAND (2 vehicles in lane_1)); C12: GREEN NS (last decision: NO_DEMAND_REST_IN_GREEN); C23: GREEN NS (last decision: NO_DEMAND_REST_IN_GREEN)

## E14: the same view at 1401 s: lane 0 still blocked, traffic still on its assigned route

![E14](../../../artifacts/final-adversarial-audit/evidence/E14.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `F/vector/95000` (scenario F, arm vector, seed 95000); sim time 1401.0 s; captured 2026-10-25T18:44:46.821943+00:00.
* **Claim it supports:** no vehicle was rerouted (route changes 0 across every run)
* **Limitation:** documents the ABSENCE of rerouting; there is no rerouting module in the SUMO path
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/F/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/F_vector_95000.jsonl.gz`
* **sha256:** `4fb9dc94d7dbc94c5a05422beadb66603d2d91016ba5e411fecbead4ca79b496`
* **State recorded at capture:** halting on approaches 160; C13: GREEN NS (last decision: NO_DEMAND_REST_IN_GREEN); C12: GREEN NS (last decision: VEHICLE_DEMAND (1 vehicles in lane_1)); C23: GREEN NS (last decision: NO_DEMAND_REST_IN_GREEN)

## E15: C41 at 1291 s, scenario I: both lanes of the C31->C41 approach blocked by vehicles stalled at the stop line since 641/662 s; VECTOR has held EW green for 43 s on a 32-vehicle queue that cannot move (it ends at max green, 45 s, and is re-served)

![E15](../../../artifacts/final-adversarial-audit/evidence/E15.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `I/vector/95000` (scenario I, arm vector, seed 95000); sim time 1291.0 s; captured 2026-10-25T18:20:46.242824+00:00.
* **Claim it supports:** VECTOR's law serves a queue that cannot discharge: halting count is its only demand signal (the mechanism of section 5.2)
* **Limitation:** one junction, one instant; the decision sequence is in the trace (E27) and the aggregate in ineffective green (S18)
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/I/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/I_vector_95000.jsonl.gz`
* **sha256:** `dc41e4ba6740e1973b7946a634718328f420e8ae179710b88912238fe0bcf03c`
* **State recorded at capture:** halting on approaches 255; C41: GREEN EW (last decision: VEHICLE_DEMAND (32 vehicles in lane_2; discharging 32)); C31: GREEN EW (last decision: VEHICLE_DEMAND (2 vehicles in lane_2)); C40: GREEN EW (last decision: VEHICLE_DEMAND (7 vehicles in lane_2))

## E16: combined stress at 1041 s: row-1 surge, C31->C41 blocked, ambulance and fire engine en route, a capacity cut due at 1100 s

![E16](../../../artifacts/final-adversarial-audit/evidence/E16.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `O/vector/95000` (scenario O, arm vector, seed 95000); sim time 1041.0 s; captured 2026-10-25T18:21:51.414795+00:00.
* **Claim it supports:** VECTOR ran the whole network under simultaneous stresses
* **Limitation:** the effect of VECTOR, if any, is in the paired metrics
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/O/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/O_vector_95000.jsonl.gz`
* **sha256:** `485c39939a69215c6ba2add926253cf4e5c7bf57ac7ebbb4b09c53dc35ad1676`
* **State recorded at capture:** halting on approaches 250; C33: GREEN NS (last decision: VEHICLE_DEMAND (1 vehicles in lane_1)); C32: GREEN EW (last decision: VEHICLE_DEMAND (1 vehicles in lane_2)); C23: GREEN NS (last decision: NO_DEMAND_REST_IN_GREEN); emergency vehicles {"amb1": {"x": 536.0, "y": 998.4, "speed": 12.74, "lane": "C14_C24_1"}, "ev1": {"x": 199.3, "y": 1001.3, "speed": 6.48, "lane": ":C04_17_0"}, "fire1": {"x": 601.6, "y": 952.5, "speed": 10.8, "lane": "C23_C24_0"}}

## E17: Primary endpoint on the held-out seeds: VECTOR minus SUMO's native actuated control, per scenario, paired 95% t-intervals; the coloured label is the unadjusted interval, the Holm column is the family-wise test (only B rejects)

![E17](../../../artifacts/final-adversarial-audit/evidence/E17.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `—` (scenario —, arm —, seed —); captured 2026-10-25T18:33:23.064Z.
* **Claim it supports:** VECTOR does not beat the native actuated controller on mean time loss; it is significantly worse in B after Holm correction
* **Limitation:** 10 paired seeds per scenario; one synthetic network; the native controller uses SUMO's own detectors
* **Linked data:** `artifacts/final-adversarial-audit/results/heldout/comparisons.json`
* **sha256:** `0aba1b3daa80d6a66796c28c6150bd4e39da68bfb4cc5b5bb4379a5343e34564`

## E18: One control cycle end to end at C21, t=1200 s, scenario A: measurement, sensor, observer, law, adapter, SUMO, shadow

![E18](../../../artifacts/final-adversarial-audit/evidence/E18.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `A/vector/95000` (scenario A, arm vector, seed 95000); sim time 1200 s; captured 2026-10-25T18:33:24.559Z.
* **Claim it supports:** the live path is SUMO -> observer -> decide_signals -> SignalAdapter -> SUMO; the shadow has no path to the decision
* **Limitation:** a single decision record; that the shadow has no influence is proven by the paired signal-sequence check, not by this image
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/A_vector_95000.jsonl.gz`
* **sha256:** `141d5874a1ff9bf8669485ab41d93057aa5602cce2173a32f297882f0ac65d8f`

## E19: Scenario M: forecasters fail at 900 s, the whole advisory orchestrator at 1500 s; control continues unchanged

![E19](../../../artifacts/final-adversarial-audit/evidence/E19.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `M/vector/95000` (scenario M, arm vector, seed 95000); captured 2026-10-25T18:33:25.111Z.
* **Claim it supports:** advisory-module failure is caught and counted, and the executed signal sequence equals VECTOR without the shadow; but after the forecasters fail the stack still reports 'degraded FULL; fallback false' with 0 forecasts, so it does not detect its own degradation
* **Limitation:** shows isolation, not graceful degradation: the failed modules never influenced control in the first place
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/M_vector_95000.jsonl.gz`, `artifacts/final-adversarial-audit/trials/heldout/M/vector/95000.json`, `artifacts/final-adversarial-audit/trials/heldout/M/vector_noshadow/95000.json`
* **sha256:** `b1da9eafe12575bcb933f1f685f7c929ebe397d65f90e4d221c5b002813d086a`

## E20: end of the combined stress run (2691 s): demand stopped at 2100 s, yet 601 vehicles are still in the network (569 halting at approaches); it did not drain

![E20](../../../artifacts/final-adversarial-audit/evidence/E20.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `O/vector/95000` (scenario O, arm vector, seed 95000); sim time 2691.0 s; captured 2026-10-25T18:23:11.914578+00:00.
* **Claim it supports:** under combined stress VECTOR leaves a residual queue at the end of the run
* **Limitation:** final metrics are in the results tables
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/O/vector/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/O_vector_95000.jsonl.gz`
* **sha256:** `3c31c958b8f865bf277141e6fdd016316f037582457843ece4a863c2bfd0cbe3`
* **State recorded at capture:** halting on approaches 569; C33: GREEN EW (last decision: NO_DEMAND_REST_IN_GREEN); C32: YELLOW NS (last decision: VEHICLE_DEMAND (13 vehicles in lane_2; max_out at 45s)); C23: YELLOW NS (last decision: NO_DEMAND_REST_IN_GREEN)

## E21: reference scenario R0 at 601 s under SUMO's native actuated baseline: the traffic state with no VECTOR and no event

![E21](../../../artifacts/final-adversarial-audit/evidence/E21.png)

* **What it is:** sumo-gui screenshot of the running simulation.
* **Run:** `R0/native_actuated/95000` (scenario R0, arm native_actuated, seed 95000); sim time 601.0 s; captured 2026-10-25T18:03:55.881273+00:00.
* **Claim it supports:** the baseline network state before any intervention or disturbance
* **Limitation:** one instant of one seed
* **Linked data:** `artifacts/final-adversarial-audit/trials/heldout/R0/native_actuated/95000.json`, `artifacts/final-adversarial-audit/traces/heldout/R0_native_actuated_95000.jsonl.gz`
* **sha256:** `9a75291eec5848fc0c71e35e16961473078273fcc594de903fdd9dcaf1136cb5`
* **State recorded at capture:** halting on approaches 45; C33: GREEN EW; C32: GREEN EW; C23: GREEN NS

## E22: Scenario E, seed 95000: ambulance, fire engine and police in every arm

![E22](../../../artifacts/final-adversarial-audit/evidence/E22.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `E/*/95000` (scenario E, arm —, seed 95000); captured 2026-10-25T18:33:25.595Z.
* **Claim it supports:** VECTOR's emergency override gives the shortest trip for all three emergency vehicles, against every baseline, on this seed and in the 10-seed means
* **Limitation:** ideal detection from simulator ground truth; ordinary vehicles do not yield; one seed shown, the 10-seed result is in emergency.json
* **Linked data:** `artifacts/final-adversarial-audit/results/heldout/emergency.json`
* **sha256:** `2ebc273701da36caacb753a98b29d2ac2cec58c776a1507dc772e4e1fa6a6ba9`

## E23: Scenario L, junction C44 (approaches swapped), t 1134-1172 s: stale (10 s delayed), contradictory, dropped and invalid observations

![E23](../../../artifacts/final-adversarial-audit/evidence/E23.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `L/vector/95000` (scenario L, arm vector, seed 95000); sim time 1134 s; captured 2026-10-25T18:33:26.688Z.
* **Claim it supports:** VECTOR acts on contradictory input without detecting it: at C44, 32-42 vehicles queue on EW while the swapped, 10 s-delayed view shows them on NS; VECTOR extends NS green until max green ends it, then serves NS again
* **Limitation:** the corruption is the audit's model; the ExtremisCanonical adapter drops the stale flag, so VECTOR cannot see it; one junction, one window
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/L_vector_95000.jsonl.gz`
* **sha256:** `1b92b662602c82e3e75dfdd8dace31e58d774a2b627f9fc945c575c682c2a8e0`

## E24: Maximum red wait: VECTOR minus native, per scenario

![E24](../../../artifacts/final-adversarial-audit/evidence/E24.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `—` (scenario —, arm —, seed —); captured 2026-10-25T18:33:24.077Z.
* **Claim it supports:** VECTOR's starvation guard (90 s) lets some approaches wait about 50 s longer than under native control, in every scenario
* **Limitation:** a worst-case statistic; it says nothing about how many vehicles waited that long
* **Linked data:** `artifacts/final-adversarial-audit/results/heldout/comparisons.json`
* **sha256:** `fb1210785ca6817007a130bbd1b0e28b0a9e612728fa734e555f3e5540ed817c`

## E25: Primary endpoint: VECTOR minus the simple vehicle-actuated rule controller

![E25](../../../artifacts/final-adversarial-audit/evidence/E25.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `—` (scenario —, arm —, seed —); captured 2026-10-25T18:33:23.560Z.
* **Claim it supports:** against a simple actuated rule VECTOR is better in several scenarios unadjusted, with no Holm rejection
* **Limitation:** secondary family; the vehicle-actuated baseline is a simple gap-out rule on the same harness
* **Linked data:** `artifacts/final-adversarial-audit/results/heldout/comparisons.json`
* **sha256:** `16b4a1db6448c13543748d683f13e93b8760bdb3e162df317e165abccbf01ff1`

## E26: Scenario K, seed 95000: three emergency vehicles in a saturated network

![E26](../../../artifacts/final-adversarial-audit/evidence/E26.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `K/*/95000` (scenario K, arm —, seed 95000); captured 2026-10-25T18:33:26.082Z.
* **Claim it supports:** under saturation the override does not guarantee arrival: on this seed the police car (pol1, E4->W1) is still in the network at the end under VECTOR, vehicle-actuated and fixed-time control; only the native programme gets it through (over 10 seeds: VECTOR 5/10, native 7/10, vehicle-actuated 0/10, fixed 2/10)
* **Limitation:** one seed shown; arrival counts over 10 seeds are in emergency.json
* **Linked data:** `artifacts/final-adversarial-audit/results/heldout/emergency.json`
* **sha256:** `4cb54e8ca439002a52ac1872c0e4cfee3a085a7b5c0cb2b3d99719e4666c0862`

## E27: Scenario I, junction C41, 1243-1320 s: every decision while both lanes of the C31->C41 approach are blocked at the stop line

![E27](../../../artifacts/final-adversarial-audit/evidence/E27.png)

* **What it is:** audit dashboard rendering recorded results.
* **Run:** `I/vector/95000` (scenario I, arm vector, seed 95000); sim time 1243 s; captured 2026-10-25T18:33:27.413Z.
* **Claim it supports:** VECTOR serves EW green to max green (45 s) on a 29-34 vehicle queue that does not discharge, then re-serves EW after clearance; its reason string labels the queue 'discharging'
* **Limitation:** one junction, one seed; the aggregate is ineffective green (S18) and the cross-cell correlation in DIAGNOSIS.md D4
* **Linked data:** `artifacts/final-adversarial-audit/traces/heldout/I_vector_95000.jsonl.gz`
* **sha256:** `19b9176ddc0358b684ce8328372651323a2aa44ec9856827b302dc92ac88813c`