Fig-5 reproduction — dose-response vs acceptance band
Increment-9 CAPSTONE — run_full_model initial viral load multiplier sweep vs targets/fig5.json
band_eval.passed at reduced scale
0 / 2 doses
reduced-scale config
1 reps · 12×12 cells
this driver's own reduced-scale run
model lethal_threshold=10000
Dose-response — final uninfected fraction vs initial viral load multiplier (log x)
Highest tested dose (10000) vs its Fig-5 acceptance band
Reduced-scale, band_eval.passed=False at every tested dose — NO reproduction claimed. `run.repro_fig5` sweeps `run_full_model` over Fig-5's initial viral load multiplier scenarios (1, 10000) and evaluates EACH dose against ONLY that dose's Fig-5 band subset (scenario-grouping guard). At this reduced-scale config (1 replica, 12×12 cells, 15 records, ~0.07 simulated days), the monotone dose-response DIRECTION holds (higher dose → not-more-surviving) but per-dose band containment fails for all 2 tested doses — expected at this scale. Mechanisms are wired and source-faithful; magnitudes are calibration-pending. Only the paper-scale ensemble (Mac-mini Phase-B follow-up) can set a `reproduced` verdict.