Guard - Leveling - Focus (look-ahead) - Summary


Convergence tuning events is the sampled subset size drawn from the preset's tuning-event pool for each convergence run.
Tail Default Convergence
Tuning events Top winner Unique winners Guard window spread Tail percentile spread Median gap Assessment
16tail-w1250-p27p5 (1/5)51250-2250 ms21.0-40.0 %0.851Noisy
24tail-w1500-p30 (2/5)41250-1500 ms30.0-40.0 %0.437Noisy
32tail-w1250-p30 (3/5)31250 ms30.0-40.0 %0.329Mixed
48tail-w1250-p30 (3/5)31250-1500 ms30.0-36.0 %0.683Mixed
64tail-w1250-p30 (1/1)11250 ms30.0 %0.718Stable
Scenario Settings
Baseline presetLeveling - Focus (look-ahead)
Target loudness-12.0 dB
Target tolerance0.0 dB
Selection rejects on low need187
Learned tail defaulttail-w1250-p30
Learned tail guard window1250 ms
Learned tail percentile30.0 %
Tuning Search
Guard window coarse values100, 150, 225, 350, 500, 750, 1000, 1500, 2250, 3000, 4500, 6000 ms
Tunable tail-percentile coarse values1.0, 2.5, 5.0, 7.5, 10.0, 15.0, 20.0, 25.0, 35.0, 50.0 %
RoundsCoarse coordinate search + single refinement round
Primary objectiveFavor real attenuation / contrast first, then preserve target-seeking and dry restoration
Fast modedisabled
Exhaustive default searchdisabled
Report events16 total (8 per split target)
Tuning events64 total (32 per split target)
Minimum upward leveling need4.0 dB at pre-event anchor
Target-tracking min work1.0 dB guard-off movement
Under-leveling thresholdseek floor < 0.35
Min tunable tail percentile1.0 % (0% kept for guard-off reference only)
Convergence study5 repeats @ 16, 24, 32, 48, 64 tuning events
Tail Clip-tuned unique configs8
Tail Configs
Learned default slugtail-w1250-p30
Learned default guard window1250 ms
Learned default tail percentile30.0 %
Clip-tuned unique configs8
Minimum upward leveling need4.0 dB
Median selected upward need5.87 dB
Clip-tuned best p25 ratio0.90
Default p25 ratio0.70
Default median seek floor1.00
Default under-leveling0.0%
Dry-relative tuning uses the raw input breath-to-anchor delta as the ideal target while retaining guard-off vs guarded attenuation and contrast metrics for continuity. Tuning candidates now require at least 4.0 dB of upward pre-event leveling need, the search is now effectiveness-first (attenuation / contrast before target-tracking tie-breakers), and 0% remains reserved for the explicit guard-off reference rather than the tunable search space. Clip-tuned best means the per-clip objective winner under the current search ordering; it is an upper-bound proxy, not a ground-truth optimum. Dry-anchor fallback was needed for 0 of 16 clips. The machine-readable tuning artifact is site/apu.software/analysis/Guard/tuning-summary.json.
Tail Guard Aggregate Metrics
Metric Learned Default Clip-tuned best
Median upward leveling need5.87 dB5.87 dB
Median guard-off pre-event target seek5.27 dB5.27 dB
P25 guard-off pre-event target seek4.76 dB4.76 dB
Median target-seek floor ratio1.000.98
P25 target-seek floor ratio0.980.87
Median target error increase0.00 dB0.00 dB
P90 target error increase0.00 dB0.00 dB
Under-leveling window rate0.0%0.0%
Applicable target-tracking windows4545
Median dry restoration ratio0.820.95
P25 dry restoration ratio0.700.90
Median dry ideal error5.29 dB1.23 dB
Median attenuation25.01 dB26.69 dB
Median contrast improvement24.80 dB26.54 dB
Median anchor drift0.02 dB0.58 dB
P90 anchor drift0.70 dB1.33 dB
Success rate100.0%100.0%
Median recovery drift0.00 dB0.00 dB
Median far-window drift0.00 dB0.00 dB
Per-clip Best-Found Settings
ClipBest-found config
dev-clean:3752_4944_000100_000013:3tail-w1250-p55
dev-clean:2086_149220_000036_000001:1tail-w1250-p55
dev-clean:5694_64029_000008_000001:1tail-w1250-p51
dev-clean:2428_83699_000015_000001:2tail-w1250-p51
dev-clean:2803_154328_000007_000001:4tail-w1250-p52p5
dev-clean:3853_163249_000064_000000:1tail-w1250-p36
dev-clean:1988_148538_000007_000000:3tail-w1250-p37p5
dev-clean:3853_163249_000160_000000:2tail-w1250-p36
test-clean:2300_131720_000003_000000:1tail-w875-p55
test-clean:4077_13754_000011_000006:1tail-w1250-p52p5
test-clean:8463_294825_000024_000000:4tail-w1250-p55
test-clean:3575_170457_000023_000002:1tail-w1250-p52p5
test-clean:7176_88083_000003_000001:1tail-w1250-p55
test-clean:4077_13754_000002_000000:1tail-w1250-p31
test-clean:4970_29095_000049_000001:2tail-w1250-p55
test-clean:4446_2275_000023_000000:2tail-w1000-p55
Tail Default Worst Dry-Ideal Misses
ClipError
test-clean:8463_294825_000024_000000:413.20 dB
test-clean:4077_13754_000011_000006:112.83 dB
dev-clean:1988_148538_000007_000000:310.44 dB
dev-clean:2428_83699_000015_000001:28.65 dB
dev-clean:3853_163249_000064_000000:18.00 dB
Tail Default Weakest Seek Floors
ClipSeek Floor
dev-clean:3853_163249_000064_000000:10.77
test-clean:4077_13754_000011_000006:10.96
test-clean:2300_131720_000003_000000:10.97
test-clean:4077_13754_000002_000000:10.97
dev-clean:3853_163249_000160_000000:20.98
Tail Clip-tuned Gain Over Default
ClipRatio Gain
test-clean:4077_13754_000011_000006:10.50
test-clean:8463_294825_000024_000000:40.47
test-clean:7176_88083_000003_000001:10.25
test-clean:4970_29095_000049_000001:20.17
dev-clean:2803_154328_000007_000001:40.16
Assertion Log 31 entries
Status Time Condition / Message
PASS 11:31:28 Guard tail default search visited 29 configs; threshold-passing=24; winner=tail-w1250-p30
PASS 11:34:24 Guard tail learned default search ordering: (0, 0, 0.0, (-1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -25.005935891255852, -18.879276243260218, -24.797921463831578, -17.545676921283224, 0.016061359462603342, 0.7009909267953534, 0.0, -5.266258685840374, -4.76257291420382, -0.9976389806216186, -0.9804176050969307, -1.0, 0.0, 0.0, -0.8200079228971597, -0.6977695644897721, 5.28695085376555, 0.0, 0.0)) (fixed baseline (0, 0, 0.0, (-0.9375, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -6.207234016657459, -2.779519736977864, -6.151497677827066, -2.7796399259322038, 0.0, 0.017665232553250476, 0.0, -5.266258685840374, -4.76257291420382, -1.0, -0.9991730792627256, -1.0, 0.0, 0.0, -0.1900626483200355, -0.11490644422539636, 22.319752541442405, 0.0, 0.0)))
PASS 11:34:24 Guard tail default median attenuation: 25.01 dB (expected >= 1.50 dB)
PASS 11:34:24 Guard tail default median contrast improvement: 24.80 dB (expected >= 1.00 dB)
PASS 11:34:24 Guard tail default median anchor drift: 0.02 dB (expected <= 0.75 dB)
PASS 11:34:24 Guard tail default p90 anchor drift: 0.70 dB (expected <= 1.50 dB)
PASS 11:34:24 Guard tail default success rate: 100.0% (expected >= 70.0%)
PASS 11:34:24 Guard tail default median recovery drift: 0.00 dB (expected <= 4.25 dB)
PASS 11:34:24 Guard tail default median guard-off pre-event target seek: 5.27 dB (expected >= 1.50 dB)
PASS 11:34:24 Guard tail default p25 guard-off pre-event target seek: 4.76 dB (expected >= 1.00 dB)
PASS 11:34:24 Guard tail default median target-seek floor ratio: 1.00 (expected >= 0.75)
PASS 11:34:24 Guard tail default p25 target-seek floor ratio: 0.98 (expected >= 0.50)
PASS 11:34:24 Guard tail default under-leveling window rate: 0.0% (expected <= 20.0%)
PASS 11:34:24 Guard tail default median target-error increase: 0.00 dB (expected <= 0.75 dB)
PASS 11:34:24 Guard tail default p90 target-error increase: 0.00 dB (expected <= 1.50 dB)
PASS 11:34:24 Guard tail clip-tuned median attenuation: 26.69 dB (expected >= 1.50 dB)
PASS 11:34:24 Guard tail clip-tuned median contrast improvement: 26.54 dB (expected >= 1.00 dB)
PASS 11:34:24 Guard tail clip-tuned median anchor drift: 0.58 dB (expected <= 0.75 dB)
PASS 11:34:24 Guard tail clip-tuned p90 anchor drift: 1.33 dB (expected <= 1.50 dB)
PASS 11:34:24 Guard tail clip-tuned success rate: 100.0% (expected >= 70.0%)
PASS 11:34:24 Guard tail clip-tuned median recovery drift: 0.00 dB (expected <= 4.25 dB)
PASS 11:34:24 Guard tail clip-tuned median guard-off pre-event target seek: 5.27 dB (expected >= 1.50 dB)
PASS 11:34:24 Guard tail clip-tuned p25 guard-off pre-event target seek: 4.76 dB (expected >= 1.00 dB)
PASS 11:34:24 Guard tail clip-tuned median target-seek floor ratio: 0.98 (expected >= 0.75)
PASS 11:34:24 Guard tail clip-tuned p25 target-seek floor ratio: 0.87 (expected >= 0.50)
PASS 11:34:24 Guard tail clip-tuned under-leveling window rate: 0.0% (expected <= 20.0%)
PASS 11:34:24 Guard tail clip-tuned median target-error increase: 0.00 dB (expected <= 0.75 dB)
PASS 11:34:24 Guard tail clip-tuned p90 target-error increase: 0.00 dB (expected <= 1.50 dB)
PASS 11:34:24 Guard tail learned default dry metrics: median ratio=0.82, p25 ratio=0.70, median error=5.29 dB, seek floor=1.00, under-leveling=0.0%
PASS 11:34:24 Guard tail clip-tuned dry metrics: median ratio=0.95, p25 ratio=0.90, median error=1.23 dB, seek floor=0.98, under-leveling=0.0%
PASS 11:34:24 Guard Leveling - Focus (look-ahead) defaults: tail=tail-w1250-p30