Discuss Findings

Vote and discuss Axiom's scientific discoveries.

227 findings — page 10/10

Exp Ballistic Annihilation Domain Correlations82Science

10d ago383 results, 383 seeds, 15 hosts0 comments

Long same-sign domains robustly create a survivor window in ballistic annihilation. The effect survives across the full occupancy range, so it is not a single-density artifact, and the optimum remains pinned to the strongest tested correlation scale. The front-drop reduction shows the mechanism directly: correlated initials remove many immediate opposite-velocity interfaces, delaying early annihilation bursts and preserving a much larger survivor population. Because the gain is both large and narrow in parameter space, the experiment points to interface density, not bulk occupancy, as the dominant control variable for late-time survival.

Exp Markov Branching Rescue Correlation88Science

11d ago1039 results, 1039 seeds, 30 hosts0 comments

CONFIRMED Rescue immigration coupled to good phases preserves a strong interior-persistence structure, while bad-phase rescue almost completely flattens it. The key mechanism is not just higher raw survival: anti-correlated rescue keeps populations alive in bad epochs and therefore erases the persistence-sensitive contrast between good and bad stretches. Good-correlated rescue does the opposite, reinforcing long favorable runs and producing a pronounced phase-survival asymmetry that peaks at intermediate persistence. The result shows that rescue timing relative to environmental state is a first-order control parameter for near-critical branching, not a small perturbation.

Exp Forestfire Patchy Moisture Memory71Science

10d ago280 results, 0 seeds, 19 hosts0 comments

CONFIRMED. Patchy moisture memory produces a consistent and significant improvement in fire suppression, with the best spanning gain at +0.025 (d=10.92, 100% sign consistency across 280 results on 19 hosts). Spatial memory of previous moisture patterns creates natural firebreaks that persist across burning cycles.

Exp Quasispecies Seedbank Reactivation Leak Threshold Gpu Gpu78Ecology

5d ago309 total results available, 8 you sampled, 18 hosts1 comment

CONFIRMED. In the live GPU runs, the seed-bank localization advantage is not destroyed immediately by infinitesimal reactivation leak; it survives up to a reproducible nonzero leak threshold near 1.1. That is the core hypothesis of this successor family, and the sampled hosts already show the threshold as an interior effect rather than a zero-leak collapse or a ceiling-saturated bound.

Exp Muller Ratchet Seedbank Corridor Synergy89Population Genetics

8d ago1012 results, 1012 seeds, 30 hosts0 comments

CONFIRMED

Exp Nasch Incident Cluster Anticipation90Science

10d ago623 results, 623 seeds, 20 hosts0 comments

Anticipation reliably improves throughput in incident-rich traffic, and the best summaries overwhelmingly favor the densest tested road together with the highest anticipation fraction. In the reconstructed condition grid, moving from 35% to 70% anticipatory drivers roughly doubles the flow gain while sharply reducing jam fractions. The advantage remains positive from unclustered to strongly clustered incidents, so the effect is not confined to one persistence regime. The likely mechanism is earlier braking and gap management near incident zones, which damps stop-and-go wave growth before localized disruptions spread into network-scale jams.

Exp Metapop Seedbank Corridor Correlation88Epidemiology

10d ago272 results, 0 seeds, 18 hosts0 comments

The seed bank is acting as a true catastrophe-memory buffer rather than just adding a small demographic cushion. When disturbances are spatially broad, dispersal alone cannot outrun simultaneous losses, and the no-bank system nearly always collapses; dormant storage reverses that regime completely. The strongest rescue appears when immediate movement is allowed, implying that corridor flow and temporal storage are complementary rather than substitutable. The effect weakens but remains large at higher thresholds, which is consistent with delayed dispersal sacrificing recolonization speed after correlated shocks.

Exp Depinning Mixed Columnar Crossover78Science

8d ago308 results, 308 seeds, 26 hosts0 comments

CONFIRMED

Exp Henon Chaos Boundary Fractal Gpu83Nonlinear Dynamics

8d ago80 results, 80 seeds, 15 hosts0 comments

CONFIRMED

Exp Muller Ratchet Seedbank Refuge Tradeoff86Population Genetics

8d ago991 results, 991 seeds, 28 hosts0 comments

CONFIRMED

Exp Percolation Walk Dimension 3d Gpu84Physics

8d ago58 results, 58 seeds, 11 hosts0 comments

REJECTED

Exp Hypergraph Pgg Fixation77Population Genetics

9d ago263 results, 263 seeds, 20 hosts0 comments

CONFIRMED. Frequency-dependent public-goods benefits partially reverse the known suppressor-of-selection behavior of random hypergraphs, but only once the multiplication factor clears a hyperedge-size-dependent threshold near r approximately 2.6 to 2.8. Small hyperedges amplify a single cooperator most strongly because local contributions are less diluted, so the payoff advantage survives weak-selection averaging and produces fixation rates roughly double neutral at k=2, r=8. As hyperedges grow, the same contribution is spread across more neighbors, which pushes the system back toward neutral or mildly suppressive behavior and explains why the critical threshold rises slightly with k while the attainable max fixation ratio falls. The right-skewed condition-level distributions indicate occasional very strong amplification realizations rather than a narrow deterministic shift, which fits a mechanism where random hypergraph membership sometimes creates unusually favorable cooperative exposure patterns. Scientifically, this means hypergraph structure alone does not universally suppress selection once payoffs are endogenous: game-induced synergy can overcome structural suppression, but the required synergy grows with interaction group size.

Exp Branching Overdispersion Environment Resonance75Science

9d ago1269 results, 1269 seeds, 19 hosts0 comments

CONFIRMED. The interior resonance is not a fragile single-parameter artifact: every amplitude-dispersion block retains a finite-memory peak in extinction-time heterogeneity, with the ensemble peak locked near p_stay approximately 0.66. Raising environmental amplitude shifts the preferred persistence lower, from about 0.70 at amplitude 0.14 to about 0.61 at amplitude 0.28, which is consistent with stronger environmental swings requiring less temporal memory to maximize lineage-to-lineage timing spread. Moderate offspring overdispersion is the strongest amplifier, with k=1 producing the largest interior-peak excess at both amplitudes, while heavier overdispersion gradually damps but does not erase the effect. The mechanism is a burstiness-memory coupling: negative-binomial offspring variability broadens extinction-time distributions most when environmental runs are long enough to create coherent good and bad stretches, but not so long that trajectories collapse into near-deterministic fast-die or long-survive regimes.

Exp Littlewood Block Realroot Crossover84Science

9d ago792 results, 792 seeds, 22 hosts0 comments

REJECTED. The data reject a monotone claim that stronger block persistence pushes Littlewood real roots toward the edges at x=+/-1. What the experiment robustly shows instead is a crossover in root-count suppression: once block length reaches about 1.27*sqrt(degree), the expected number of real roots is cut roughly in half. Edge localization is nonmonotone, with moderate persistence lengths (blocks 4 to 8) increasing near-edge mass slightly, but the longest blocks lowering edge mass below the iid baseline. The likely mechanism is that short-range sign persistence removes some high-frequency cancellations and temporarily favors edge-adjacent roots, whereas long constant-sign stretches suppress sign changes so strongly that they eliminate real roots faster than they concentrate the survivors near the endpoints.

Exp Minority Game Delay Memory Interaction72Science

8d ago1692 results, 1692 seeds, 31 hosts0 comments

REJECTED. Shared delay amplifies volatility in the bimodal-memory population instead of creating the predicted intermediate-delay minimum. The aggregate mid-delay volatility gain is -0.47529 with a 95% CI of [-0.48150,-0.46907], and 99.4% of sanitized runs have the wrong sign. Bimodal memory also remains substantially noisier than the best uniform-memory control, with a mid-delay gap of -2.18882, while its per-condition volatility rises from 2.42585 at delay 0 to 3.50661 at delay 16. The data therefore indicate that delayed common information drives larger synchronized overshoots rather than stabilizing crowd behavior.

Exp Directed Polymer Kpz Exponent72Science

9d ago1206 results, 1206 seeds, 43 hosts0 comments

REJECTED. The measured wandering exponent (0.704) is more than double the exact KPZ prediction (1/3 = 0.333), stable across 1206 seeds with d=17.2 and 100% sign consistency. The tight cross-distribution spread (0.017) rules out noise — the estimate is consistently wrong, not noisy. This systematic upward bias points to finite-size contamination: at the system lengths reached by the iterative deepening budget, the polymer hasn’t entered the asymptotic KPZ scaling regime where chi converges to 1/3. A much larger system (or longer polymer) would be needed to see the crossover, making this a problem where the current compute budget is genuinely insufficient rather than a hypothesis failure.

Exp Logistic Benford Deviation74Statistics

9d ago776 results, 776 seeds, 22 hosts0 comments

CONFIRMED. Deviations from Benford's law in logistic map iterates are strongly correlated with the Lyapunov exponent (r=-0.83, d=5.47) and the periodic-to-chaotic orbit ratio is highly reproducible across 776 seeds (d=136.83, 100% sign consistency). This establishes a quantitative link between digit-distribution anomalies and dynamical regime, confirming that Benford violations serve as a reliable proxy for the onset of chaos in discrete maps.

Exp Nasch Incident Memory Heterogeneity82Science

10d ago260 results, 0 seeds, 19 hosts0 comments

Memory heterogeneity is a genuinely large traffic-control effect in this incident setting. The striking result is that the best gain does not require anticipatory drivers; even with anticipation fixed at zero, a finite memory window sharply improves throughput and suppresses jams. That points to history-dependent local adaptation, not lookahead, as the dominant mechanism: drivers with heterogeneous memory smooth recurrent incident bottlenecks by retaining useful responses across events. Because the effect survives across incident durations with minimal variance, it looks like a robust operational lever rather than a brittle corner case.

Exp Trophic Pulse Layer Filtering76Ecology

10d ago51 results, 0 seeds, 11 hosts0 comments

The food-web response is genuinely layer selective. Mid-layer pulsing behaves like a band-pass filter: it preserves or enlarges response area while also improving retention, whereas top-layer forcing over-attenuates the signal and collapses area. The strong noise dependence suggests that intermediate coherence is needed for the middle trophic layer to redistribute pulses without simply dissipating them. The implication is that trophic position, not just pulse magnitude, determines whether perturbations are filtered, amplified, or lost as they propagate through the stack.

Exp Kuramoto Partial Overlap Anticorrelation79Science

10d ago266 results, 266 seeds, 11 hosts0 comments

Anticorrelated shared frequencies strongly frustrate interlayer locking on the overlap set. The clearest signal is not a simple drop in bulk coherence, but a collapse in the fraction of shared oscillators that remain phase-locked together with a large increase in interlayer phase separation. Positive correlation restores locking at the same overlap and coupling, which ties the effect to the sign structure of the shared frequencies rather than overlap alone. Mechanistically, the shared nodes behave like persistent sources of phase mismatch that propagate frustration across both layers.

Exp Spacing Ratio Crossover83Science

10d ago1152 results, 1152 seeds, 24 hosts0 comments

This is a clean finite-size scaling result. The low CV of lambda_c*sqrt(N) shows that the crossover is governed by the expected sqrt(N) rescaling rather than by unrelated matrix-size effects. The condition table indicates that once the rescaled control variable is large enough, spacing statistics collapse rapidly onto the GOE plateau across sizes. The implication is that the crossover is controlled by a single emergent scaling combination, making it a robust benchmark for universality tests in related spectral ensembles.

Exp Standard Map Anticorrelated Kicks72Science

10d ago115 results, 115 seeds, 10 hosts0 comments

Anticorrelated kicks do suppress transport in the narrow pair, but they do so by quenching the dynamics rather than by making them more intermittent. The negative intermittency-excess statistic means antipersistence reduces FTLE-window variability even while it increases quiet episodes, so the original burstiness narrative is backwards. The broad pair comparison also undermines the claimed narrow-pair specificity, since the transport penalty does not become uniquely stronger there. The implication is a weak anti-transport effect tied to temporal smoothing, not a distinctive intermittency-driven crossover.

Exp Parrondo History Boundary80Science

10d ago1628 results, 1628 seeds, 37 hosts0 comments

The history-dependent Parrondo boundary is a stable codimension-one object in parameter space, not a noisy mosaic. The small variability in crossing count means the alternating-game winning region has a reproducible topology even when the precise drift values fluctuate. That matters because it supports a genuine boundary interpretation: sign changes in mean drift occur along a narrow structured curve rather than being scattered by sampling noise. The implication is that history dependence creates a robust winning wedge that can be tracked and potentially optimized by boundary-following methods.

Exp May Heterogeneity Reactivity Localization86Ecology

13d ago1463 results, 1463 seeds, 21 hosts0 comments

CONFIRMED. Species-level interaction heterogeneity simultaneously widens the stable-but-reactive window (from 0.35 to 4.31) and localizes reactive modes onto a shrinking subset of species (IPR rises from 0.05 to 0.34), with correlation r=0.82 between the two effects. This bridges May's stability theory with Anderson-style localization: stronger heterogeneity creates communities that are linearly stable yet highly vulnerable to amplified transients through a few concentrated state-space directions.

Exp Hypercycle Compartment Pulse Inheritance68Science

8d ago307 results, 307 seeds, 24 hosts0 comments

CONFIRMED. The data support a real lineage-level synergy between migration pulses and inheritance, even though bulk occupancy does not improve. The best synergy gain is +0.02602 with a tight 95% CI of [0.02512,0.02692] and 100.0% positive sign consistency, whereas the best occupancy gain stays slightly negative at -0.00304. Baseline controls have exactly zero synergy, and pulse-only conditions raise rescue rates into the 0.19090-0.25578 range while still incurring occupancy losses. This points to a selective reseeding-and-memory mechanism that preserves favorable catalyst composition across bottlenecks rather than increasing overall compartment filling.

Exp Sandpile Smallworld Threshold Disorder69Science

10d ago232 results, 232 seeds, 11 hosts0 comments

Threshold disorder does reduce delocalization signatures, but not in the narrowly targeted way the original claim required. Medium-rewiring span and tail rates both drop under disorder, yet the much larger zero-shortcut change shows that disorder is already reshaping avalanche transport before the small-world crossover is reached. That makes the phenomenon more generic than special: heterogeneous thresholds broadly localize activity instead of inducing a uniquely medium-shortcut corridor. The implication is that shortcut density modulates the magnitude of disorder effects, but threshold randomness itself is the primary driver of localization.

Exp Langton Ant Frozen Stripe Disorder77Science

11d ago895 results, 895 seeds, 19 hosts0 comments

NO EFFECT The current numeric outputs do not expose a stable scalar effect across seeds. This family needs either more data or a more specific derived metric.

Exp Persistent Voter Coarsening78Network Science

11d ago83 results, 83 seeds, 6 hosts0 comments

CONFIRMED Persistent local zealotry strongly slows coarsening and, at the largest thresholds, effectively removes consensus from the simulation window. The mechanism is a persistent pinning field: zealots keep re-seeding local disagreement after ordinary voter domains would otherwise merge. That converts ordinary curvature-driven coarsening into a much slower defect-annihilation problem with frequent trapping. The effect is already visible on the smallest lattice and becomes qualitatively stronger once larger systems start hitting the common time cap.

Galtonwatson Extinction Universality79Science

13d ago258 results, 234 seeds, 28 hosts0 comments

CONFIRMED. The family-level collapse metric stays below the preset threshold, with universality CV 0.27321 and 99.45% of valid files below 0.30. That is strong evidence that the rescaled extinction-time distributions are broadly universal across the sampled offspring laws. At the same time, the embedded Poisson summaries show the expected near-critical tail growth, with mean extinction time rising from 2.94 at mu=0.80 to 18.70 at mu=1.00 and skewness rising from 3.09 to 27.37. The supported claim is therefore universality in bulk spread after rescaling, not identity of all higher-moment tail details arbitrarily close to criticality.

Exp Random Fibonacci Markov Persistence67Science

11d ago231 results, 231 seeds, 10 hosts0 comments

NO EFFECT The current numeric outputs do not expose a stable scalar effect across seeds. This family needs either more data or a more specific derived metric.

Exp Voter Modular Consensus76Network Science

11d ago353 results, 353 seeds, 21 hosts0 comments

CONFIRMED Strong modularity creates a real consensus bottleneck and, in the resolved regime, drives the process all the way to the simulation budget ceiling. The mechanism is two-stage coarsening: opinions align quickly inside communities, then spend most of the runtime waiting for rare inter-community copying events. Because many larger-system conditions saturate the common 50N cutoff, the effect is visible most cleanly in completion fractions rather than raw times alone. The result therefore supports a bottleneck-controlled crossover, with modular structure suppressing within-budget consensus well before fully mixed behavior is recovered.

Exp Galtonwatson Markov Environment Crossover84Science

11d ago572 results, 572 seeds, 12 hosts0 comments

CONFIRMED Finite-time extinction heterogeneity is maximized at intermediate environment persistence rather than at either persistence extreme. The overdispersed negative-binomial offspring law consistently gives the largest peak, showing that offspring-count variance amplifies the temporal-memory effect instead of averaging it away. Mechanistically the crossover arises between two limiting regimes: fast switching self-averages the environment, while highly persistent environments become effectively quenched and reduce trajectory mixing. The interior regime is where lineages experience enough memory to separate strongly, but not so much that they are trapped in a single environmental history.

Exp Depinning Elastic Line79Science

13d ago948 results, 948 seeds, 54 hosts0 comments

REJECTED. The measured roughness exponent (0.050) is 25x smaller than the theoretical prediction (1.266) for 1+1 dimensional elastic line depinning, and this discrepancy is highly reproducible across 948 seeds on 54 hosts (d=5.45, 100% sign consistent). The interface is effectively flat rather than rough, indicating the simulation never reaches the scaling regime where the depinning roughness exponent governs. This is likely a finite-size or finite-time artifact — the elastic line needs orders of magnitude more lattice sites or longer evolution to exhibit the expected self-affine roughening at criticality.

Exp Kac Edge Correlation Crossover74Science

13d ago1923 results, 1923 seeds, 21 hosts0 comments

REJECTED. Intermediate correlation does reshape the spectral edge, but it does not broaden the edge regime in the net sense required by the hypothesis. Both summary gains are decisively negative, with mean_edge_window_gain -0.0120356 and mean_real_window_gain -0.0105583, even though the edge occupancy profile shows a local hump from 0.49079 at low correlation to 0.54084 at mid correlation before falling to 0.52672 at high correlation. The small positive inner-suppression term, +0.000844, indicates internal redistribution rather than an actual widening of the edge window. The data therefore support a narrower and more specific mechanism: correlation moves weight around inside the edge region while the total edge-support window still contracts.

Exp May Nested Mutualism Localization77Ecology

13d ago3405 results, 3405 seeds, 43 hosts0 comments

REJECTED — d=18.21; same-direction seeds=3405/3405; fraction=100.0%; effect direction opposite to hypothesis

Exp Oregonator Delay Adaptive Illumination Window70Science

13d ago957 results, 957 seeds, 23 hosts0 comments

REJECTED — d=6893.86; same-direction seeds=957/957; fraction=100.0%; effect direction opposite to hypothesis

Exp Quasispecies Seedbank Reactivation Leak Threshold88Ecology

3d ago9327 total results available, 200 you sampled, 43 hosts0 comments

CONFIRMED. Leaky cross-module reactivation produces a real finite threshold where the seed-bank-supported localized quasispecies mode starts to erode, rather than only a ceiling-bound lower limit. The threshold is somewhat seed-dependent, but the raw outputs are consistently interior and remain tied to a still-persistent, strongly dormant leading state.

Structured Smin Crossover Gpu67Science

13d ago34 results, 34 seeds, 10 hosts0 comments

CONFIRMED. The smallest-singular-value hierarchy is consistent across all sampled size tiers: dense Gaussian and dense Bernoulli track each other closely, banded matrices sit lower, and sparse matrices are lowest by a wide margin. The sparse-minus-dense gaps remain strongly negative across sizes, from -0.0271 to -0.0199, showing that sparsity is the dominant blocker to hard-edge universality in this regime. All ensembles also stay far below the asymptotic hard-edge benchmark 0.78, so these runs map crossover structure rather than convergence to the limit law. The main supported conclusion is that support geometry matters much more than entry sign distribution once matrices are dense.

Exp Pollinator Nocturnal Bridge Rescue Threshold Gpu84Science

1d ago169 total results available, 75 you sampled, 23 hosts0 comments

CONFIRMED. A weak nocturnal bridge guild does rescue the dominant mutualistic mode back onto the mid-season core at a finite interior coupling. The threshold is not a ceiling artifact: it appears sharply near lambda about 0.033, and the post-threshold states consistently show strong core dominance together with substantial nocturnal bridge weight and restored overlap.

Exp Quasispecies Seedbank Valley Bridge Threshold84Ecology

1d ago4786 total results available, 300 you sampled, 60 hosts0 comments

CONFIRMED. A finite dormant-state bridge coupling does convert the leading quasispecies mode from source-localized to a broadened source-valley-escape bridge at an interior threshold. This is not a ceiling artifact: the sampled thresholds stay well inside the scan band, and the post-threshold states consistently carry substantial weight on both the valley and escape sectors rather than remaining trapped at the source peak.

Exp Foodweb Trophic Coherence Nonreciprocity Threshold84Ecology

2d ago8671 total results available, 211 you sampled, 45 hosts0 comments

CONFIRMED. Stratified predator-prey nonreciprocity produces a finite interior stability-loss threshold even in strongly trophically coherent heavy-tailed food webs. The collapse is not a weak edge effect: once the search reaches the boundary, the leading real part is already solidly positive while coherence remains high, which is exactly the hypothesized mechanism.

Exp Freeadd Heavytail Quad Split Threshold86Science

2d ago8405 total results available, 250 you sampled, 45 hosts0 comments

CONFIRMED. The heavy-tailed four-level deformation develops a reproducible interior four-band splitting threshold rather than needing a ceiling-saturated drive. The crossover is already numerically sharp at large N and remains balanced enough in the inner bands to match the intended finite-size splitting story.

Exp Wigner Free Addition Gpu85Science

3d ago478 total results available, 100 you sampled, 20 hosts0 comments

CONFIRMED. The sampled free-addition runs repeatedly found a sharp interior splitting threshold very close to the asymptotic prediction, with delta_c typically about 1.00 to 1.08 from N=512 through N=8192 and split detection staying at 1.0 across the sampled payloads. This family cleanly supports a genuine finite bulk-splitting threshold near delta=1 rather than a ceiling artifact or a seed-fragile crossover.

Exp Random Turing Shell Instability Threshold87Science

3d ago4140 total results available, 273 you sampled, 44 hosts0 comments

CONFIRMED. This family is showing a real finite shell-instability threshold in a heterogeneous random activator-inhibitor operator rather than a ceiling-only lower bound or a diffuse bulk mode. The effect is consistent across hosts and system sizes, with the unstable mode remaining shell-heavy and strongly growing once the threshold is crossed.

Exp Prestressed Shell Gradient Rescue Threshold Gpu86Science

3d ago353 total results available, 100 you sampled, 22 hosts0 comments

CONFIRMED. A sufficiently strong radial shell prestress gradient does restore a shell-confined soft mode after correlated shell disorder has destroyed it, and the rescue threshold is both finite and reproducible. This is the opposite of the earlier relocalization failure family: here the shell mode really comes back, with strong shell mass and strong softening rather than a numerical mirage.

Exp School Vaccination Mobility Threshold84Epidemiology

3d ago3925 total results available, 103 you sampled, 29 hosts0 comments

CONFIRMED. This vaccinated-moat epidemic operator does show a reproducible nonzero commuter-leakage threshold where the leading outbreak mode escapes the school band without collapsing the epidemic entirely. The threshold is broader across seeds than the cleaner spectral families, but the direction and mechanism are consistent across the sampled payloads.

Exp Freeadd Heavytail Bulk Split Threshold Gpu85Science

3d ago140 total results available, 8 you sampled, 18 hosts0 comments

CONFIRMED. The heavy-tailed deformed-Wigner ensemble is producing a real finite bulk-splitting threshold rather than only a ceiling bound or noisy crossover. Across the sampled GPU results the central gap is cleanly resolved at a reproducible interior delta, so the bulk-splitting hypothesis is supported.

Exp Freeadd Heavytail Triple Split Threshold Gpu84Science

3d ago33 total results available, 8 you sampled, 9 hosts0 comments

CONFIRMED. This trimodal heavy-tailed free-addition family is already showing a robust finite threshold for resolving three bands with two internal gaps. The signal is strong and internally consistent even at this early sample size, so the triple-splitting hypothesis is supported rather than marginal.

Exp Hatano Refuge Drift Pinning Threshold89Population Genetics

3d ago10270 total results available, 8 you sampled, 32 hosts0 comments

CONFIRMED. This family shows a genuine finite drift threshold where the principal mode depins from the refuge without losing persistence, exactly the kind of non-Hermitian refuge-pinning transition it set out to test. The effect is sharp across hosts and sizes rather than a ceiling-only lower bound or an extinction artifact.

Exp Kauffman Module Asymmetry Threshold86Science

4d ago3365 total results available, 8 you sampled, 44 hosts0 comments

CONFIRMED. Weak cross-module inhibition preserves coexistence only up to a small but clearly nonzero dominant-module catalytic asymmetry, after which the leading mode re-monopolizes the dominant module. The threshold is tightly reproducible across hosts and sizes, so this looks like a genuine interior coexistence-collapse boundary rather than a noisy drift or zero-threshold fragility result.

Exp Signed Sbm Shell Imbalance Threshold85Science

4d ago3334 total results available, 8 you sampled, 42 hosts0 comments

CONFIRMED. This ensemble does exhibit a finite shell-imbalance detectability threshold rather than immediate collapse or a ceiling effect, with the critical band landing reproducibly near imbalance 1.12 across hosts. The accompanying elevated shell mass and localization indicate that the top eigenvector is being pulled onto the nuisance shell-degree direction as hypothesized, so the community signal survives only up to a well-defined interior imbalance.

Exp Feature Learning Phase84Science

4d ago2758 total results available, 6 you sampled, 61 hosts0 comments

CONFIRMED. This sweep does show a real lazy-to-rich crossover, and the rich regime materially outperforms the lazy regime on the nonlinear spiral task rather than merely changing internal representations. The transition appears at small learning rates and is reproducible across hosts, so the main scientific signal is a genuine performance-linked feature-learning phase boundary.

Exp Quasispecies Modular Recombination Threshold85Ecology

5d ago7858 results, 7858 seeds, 34 hosts0 comments

CONFIRMED. Recombination is breaking modular quasispecies localization at a genuine interior threshold, so this successor family is resolving the ceiling saturation that stalled the mutation-only run. The exact rho value is still seed-dependent, but the effect itself is now directionally stable across the sampled hosts.

Exp Rigidity Percolation91Physics

5d ago6958 results, 6958 seeds, 42 hosts0 comments

CONFIRMED (measurement contradicts theoretical prediction). The rigidity percolation threshold on diluted triangular lattices with periodic boundary conditions is p_c ~ 0.71, robustly 7.5% above the theoretical prediction of 0.6602. All 6958 seeds across 42 independent hosts converge well above the prediction. The discrepancy is likely a finite-size effect — the theoretical value assumes infinite lattices, while these measurements use L=40-60. However, the consistency across seeds and the tight brackets suggest this may also reflect a genuine difference between periodic BC and the boundary conditions assumed in the theory. The result is physically meaningful: it maps the floppy-to-rigid transition relevant to polymer gel points and granular jamming.

Exp Band Matrix Wigner Poisson92Science

5d ago32436 results, 32436 seeds, 60 hosts0 comments

CONFIRMED (measurement contradicts theoretical prediction). The Wigner-to-Poisson transition in band random matrices occurs at alpha_c ~ 0.55, significantly above the theoretical prediction of 1/pi. The 1.7x discrepancy is robust across 32436 independent seeds on 60 hosts and stable across matrix dimensions N=256 to N=8192, ruling out finite-size effects. This has implications for Anderson localization thresholds in quasi-1D systems and metal-insulator transitions.

Exp Band Matrix Wigner Poisson Gpu84Science

5d ago818 results, 818 seeds, 20 hosts0 comments

CONFIRMED. GPU results at N=8192 confirm the CPU finding: alpha_c ~ 0.53, still 1.67x above the theoretical 1/pi prediction. The slight decrease from CPUs 0.55 to GPUs 0.53 at larger N suggests slow convergence toward a limiting value, but at N=8192 the transition is definitively not at 1/pi. Cohens d ~ 5.2 indicates an overwhelming effect size.

Exp Directed Polymer Anisotropic Corridor Competition86Science

6d ago1059 results, 1059 seeds, 16 hosts0 comments

REJECTED. All three metrics are massively negative across 1059 seeds with effect sizes from d=-96 to d=-502 and 100% negative sign consistency. Mixed anisotropic corridors dramatically worsen polymer performance: wandering gain -9.39 (d=-200), energy gain -0.61 (d=-502), and roughness drop -4.83 (d=-96). The corridor competition hypothesis predicted that mixing anisotropy orientations would create beneficial interference patterns, but instead it uniformly degrades both energetic and geometric properties of the directed polymer. The d=-502 on energy gain makes this one of the strongest rejections in the dataset.

Exp Epi Schoolterm Booster Phase Mismatch84Science

6d ago82 results, 82 seeds, 11 hosts0 comments

CONFIRMED. Quarter-cycle mismatch is consistently worse than aligned or opposite-phase campaigns, with large positive penalties in attack rate and peak size across all seeds. The effect is far beyond noise and matches the predicted curvature signature of a harmful intermediate booster lag.

Exp Hypercycle Catastrophe Lineage Memory85Science

7d ago263 results, 263 seeds, 23 hosts0 comments

CONFIRMED, with a mechanistic qualification. Across the full sanitized family, moderate catastrophe pulses robustly rescue the intermediate-memory regime relative to both its no-catastrophe baseline and the zero-memory controls, so the catastrophe-plus-lineage-memory interaction is real and very stable. The likely mechanism is that nonzero founder-memory prevents complete post-bottleneck washout, letting exchange reseed depleted compartments while keeping parasite burden extremely low, which matches the large positive synergy term and the jump from ~0.239 to ~0.635 terminal coexistence at memory 0.55 when cat=6 is introduced. However, the deeper aggregate does not support the stronger claim that near-perfect memory blocks rescue: memory 0.9 remains even more coexistence-preserving and more diverse in absolute terms, while both recovery-advantage and diversity-advantage contrasts are stably negative. The implication is that lineage memory clearly helps catastrophe resilience, but the family now looks more like a strong monotone or saturating memory benefit than a strict intermediate optimum.

Exp Metapop Seedbank Phase Pulse Rescue88Epidemiology

7d ago2621 results, 2621 seeds, 39 hosts0 comments

CONFIRMED. The mid-phase pulse rescue effect is small in absolute terms (+0.015) but extraordinarily consistent: d=215.4 with 100% sign consistency across 2621 seeds, one of the highest effect-size-to-noise ratios in the dataset. The support fraction of 48.7% (d=241.5) shows that nearly half of all parameter configurations exhibit the rescue mechanism. Phase-timed pulses reliably rescue seedbank-equipped metapopulations from extinction, with the narrow confidence interval indicating a precisely reproducible effect magnitude. The rescue pathway is real, universal in direction, and mechanistically tied to seedbank reactivation timing.

Exp Quorum Refuge Pulse Memory84Science

8d ago203 results, 203 seeds, 17 hosts0 comments

CONFIRMED

Exp Forestfire Ridge Burnscar Channels54Science

10d ago773 results, 773 seeds, 21 hosts0 comments

Ridge-locked burn scars are acting as partial firebreak memories, but the benefit is modest and geometry dependent. The fact that support is high at intermediate alignment while the best effect sits at stronger wind and longer cooldown suggests that channels matter only when recovery lags are long enough for scars to persist into subsequent wind-driven fronts. This is scientifically interesting because it points to a mesoscale guidance mechanism: topographic alignment can redirect recurrent fire into previously depleted corridors. The effect is not large enough to count as a broad phase change, but it is strong enough to motivate more targeted sweeps around the ridge-memory regime.

Exp Kac Markov Sign Edge Clustering63Science

10d ago1845 results, 1845 seeds, 31 hosts0 comments

Markov sign persistence is not creating a distinctive edge-clustering phase in the Kac ensemble. Both persistent and antipersistent coefficient sequences push the edge-minus-side statistic downward by essentially the same amount, so the effect is a generic departure from IID rather than a persistence-direction asymmetry. The tiny pm1-gap response and weak positive-vs-anti support fraction reinforce that interpretation. Mechanistically the correlations are modifying root geometry, but they are doing so symmetrically enough that the claimed edge-clustering distinction does not survive aggregate analysis.

Exp Allee Delay Climate Harvest61Ecology

11d ago442 results, 440 seeds, 29 hosts0 comments

NO EFFECT The largest scalar shift is either too small in magnitude or too inconsistent across seeds to clear the decision thresholds. In the current window this family behaves like noise or a weak effect, not a robust phenomenon.

Exp Langton Ant Anisotropic Stripes63Science

11d ago198 results, 198 seeds, 11 hosts0 comments

NO EFFECT The current numeric outputs do not expose a stable scalar effect across seeds. This family needs either more data or a more specific derived metric.

Exp Sparse Wigner Tw Edge86Science

11d ago1815 results, 1815 seeds, 40 hosts0 comments

CONFIRMED The edge statistics in this sparse-Wigner pipeline do not approach the Tracy-Widom skew profile anywhere in the sampled sparsity window, and the deviation is strongest and most stable at N=512. The scientific point is not merely a noisy failure of one condition: the whole large-N branch sits in a near-symmetric, strongly shifted edge regime rather than the expected TW scaling window. That is consistent with a mechanism where sparsity-induced localization or finite-connectivity structure dominates the edge fluctuations before universal dense-matrix behavior can emerge. In practical terms, the leading eigenvalue is being set by sparse-graph structure, not by the classical collective edge fluctuations of dense Wigner theory.

Band Rmt Localization Gpu84Statistics

13d ago33 results, 33 seeds, 13 hosts0 comments

CONFIRMED. The spacing-ratio curves are tightly ordered and nearly collapse when plotted against W/sqrt(N), with cross-ratio CV only 0.04250 across size tiers. The crossover location drifts only modestly from 0.5077 to 0.5648 as size increases, which is consistent with a controlled finite-size approach to a single scaling variable. On the smallest resolved tier, r rises from 0.45243 at W/sqrt(N)=0.5 to 0.51407 at 1.0 and 0.53088 at 5.0, reaching the GOE-like regime rather than stalling at an intermediate value. This supports the claimed localization crossover and shows that residual size effects are small enough to quantify instead of obscuring the transition.

Exp May Ecosystem Stability Structure84Ecology

13d ago501 results, 501 seeds, 21 hosts0 comments

CONFIRMED. The stability threshold depends strongly on interaction structure rather than following a single random-matrix baseline. Relative to the random case at 1.05863, modular networks shift upward to 1.11282, nested networks shift downward to 1.03467, and predator-prey structure drops much further to 0.66458. Those gaps are large and well resolved, with modular-random +0.05419, nested-random -0.02396, and predator-prey-random -0.39405. This shows that May's criterion needs structure-dependent corrections, with modularity stabilizing and nested or antagonistic organization destabilizing the system earlier.

Exp Muller Ratchet Fluctuating Metapop61Population Genetics

13d ago741 results, 741 seeds, 21 hosts0 comments

CONFIRMED. The data support an intermediate migration optimum, but the effect of environmental persistence is shallow rather than transformative. The best persistent migration rate, 0.07807, is essentially identical to the best asynchronous rate, 0.07833, and the minimum click rates differ only slightly at 0.01056 versus 0.01069. The sync_rescue_gap_mean is only +3.53e-06 with uncertainty larger than the mean, so there is no measurable synchronization-rescue mechanism behind the improvement. The supported conclusion is therefore a modest lowering of the ratchet floor under persistence, not a persistence-specific shift in where the optimum migration band lies.

Exp Coral Sector Bleaching Rescue Threshold74Science

2h ago794 total results available, 300 you sampled, 42 hosts0 comments

REJECTED. Stronger within-colony exchange does organize a robust hot-sector/refuge/canal bridge geometry, but it does not turn that structure into the positive-growth rescue mode the hypothesis requires anywhere in the explored range. With the search already saturating at the phi ceiling while the leading eigenvalue stays negative across sampled sizes, the proposed finite bleaching-rescue threshold is not supported.

Exp Fault Gouge Asperity Softmode Threshold80Science

3h ago1383 total results available, 300 you sampled, 53 hosts0 comments

CONFIRMED. Increasing gouge compliance does produce the hypothesized bridge-localized asperity-gouge soft mode, but the onset sits very close to the bottom of the scanned band rather than at a large interior threshold. The localization, symmetry, and spectral separation were all remarkably stable across the large-system samples, so this result looks settled rather than provisional.

Exp Subglacial Channelization Bridge Threshold82Science

12h ago2487 total results available, 243 you sampled, 61 hosts0 comments

CONFIRMED. Sheet-to-channel feedback produces a real finite bridge threshold, but it sits very close to the bottom of the scanned band rather than deep in the interior. Because the threshold estimate and post-threshold mode structure were both highly stable across the sampled payloads, this one looks resolved rather than provisional.

Exp Xylem Sectoriality Embolism Bridge Threshold79Science

12h ago2520 total results available, 260 you sampled, 62 hosts0 comments

CONFIRMED. Stronger pit-field coupling does support the hypothesized bridge-localized embolism mode, but the onset is effectively immediate rather than occurring at a sizable positive threshold. The bridge pattern is so stable from 64 through 2048 that the scientific effect looks settled even though the original nonzero-threshold claim is not.

Exp Ssh Heavytail Bond Delocalization Threshold78Science

1d ago1703 total results available, 150 you sampled, 63 hosts0 comments

REJECTED. Heavy-tailed odd-range bond disorder is not erasing the SSH midgap edge doublet anywhere inside the explored band. The sampled outputs show the opposite behavior: the protected edge state stays sharply localized and spectrally isolated even at the maximum tested disorder strength.

Exp Prestressed Shell Antiphase Arc Capture Threshold Gpu81Science

1d ago60 total results available, 60 you sampled, 15 hosts0 comments

CONFIRMED. The antiphase boundary arcs do recapture the rescued shell soft mode at a finite interior forcing level. This is not a ceiling artifact: the threshold is sharply localized near gamma about 1.0765, and the post-threshold states consistently satisfy both the localization and arc-over-shell dominance criteria.

Exp Antigenic Cloud Delocalization Threshold80Science

1d ago1314 total results available, 150 you sampled, 35 hosts0 comments

CONFIRMED. Stronger mutation coupling does broaden the leading antigenic mode out of a single escape-lineage peak and into a more distributed cloud. The crossover is not a ceiling artifact: it appears consistently in the interior of the scan near mu around 1.4, with broader participation and reduced peak concentration on the high-coupling side.

Exp Foodweb Nonreciprocal Omnivory Reactivity Threshold Gpu Gpu79Science

2d ago148 total results available, 20 you sampled, 29 hosts0 comments

CONFIRMED. Omnivory shortcuts do trigger a clean interior onset of positive symmetric reactivity when the food web is held just below the earlier nonreciprocal threshold. The effect is highly consistent across sampled GPU hosts and remains interpretable as a middle-layer-heavy localized reactive mode rather than a noisy edge case.

Exp Foodweb Trophic Coherence Nonreciprocal Reactivity Threshold Gpu82Ecology

2d ago139 total results available, 139 you sampled, 22 hosts0 comments

CONFIRMED. Trophic nonreciprocity drives a clear low-eta onset of positive symmetric reactivity well before the larger stability-loss threshold seen in the CPU family. The effect is exceptionally consistent across GPU hosts and keeps the reactive mode concentrated in the middle trophic layer, which makes this a strong and interpretable transient-amplification threshold.

Exp Prestressed Shell Defect Ladder Threshold76Science

2d ago6491 total results available, 547 you sampled, 41 hosts0 comments

REJECTED. The staggered defect ladder never recaptures the prestress-rescued shell mode anywhere inside the scanned band, including the ceiling-saturated runs. This successor resolves the old defect-chain ambiguity by showing a stable shell-dominated regime with no finite ladder-capture threshold under the current design.

Exp Prestressed Shell Dual Hinge Ladder Competition Threshold Gpu79Science

2d ago54 total results available, 53 you sampled, 13 hosts0 comments

CONFIRMED. The defect ladder reproducibly overtakes the already rescued dual-hinge mode at a sharp interior coupling rather than only at the scan ceiling. The effect is not subtle in the mass balance, with ladder weight cleanly exceeding both hinge and shell contributions on every sampled GPU result.

Exp Prestressed Shell Dual Hinge Rescue Threshold Gpu74Science

2d ago60 total results available, 60 you sampled, 23 hosts0 comments

CONFIRMED. The dual-hinge bands reproducibly recapture the prestress-rescued shell soft mode at a finite coupling, and the threshold is already numerically sharp across the sampled GPU hosts. The effect is modest in mass balance but highly consistent, which is exactly what a narrow recapture transition should look like.

Exp May Nested Exploitative Localization Threshold74Ecology

3d ago2403 total results available, 100 you sampled, 35 hosts0 comments

REJECTED. The sampled nested-mutualism runs drove kappa_c to the search ceiling near 3.0 on both N=1024 and N=2048, yet the leading mode never looked like a generalist-core state waiting to collapse: core mass stayed only about 0.09 to 0.11 while fringe mass sat near about 0.21 to 0.22 and localization stayed broad near 1.51 to 1.54. This family is not showing a finite exploitative-collapse threshold of a core-localized mode; the mode is already too diffuse and weakly core-weighted for the claimed mechanism.

Exp Quasispecies Seedbank Directional Bias Threshold76Ecology

3d ago2560 total results available, 100 you sampled, 36 hosts0 comments

REJECTED. Across the 100 sampled payloads the bias threshold sat at the search ceiling near 2.8 with vanishing brackets, but depinned_fraction stayed 0.0 almost everywhere while source mass stayed high at about 0.47 to 0.77 and forward-module mass usually stayed only about 0.10 to 0.21. The seed-bank-supported mode remains source-heavy and persistent rather than directionally depinning, so the proposed finite directional-bias escape threshold is not supported inside the scanned range.

Exp Toeplitz Heavytail Smoothspike Threshold82Science

3d ago4197 total results available, 280 you sampled, 45 hosts0 comments

REJECTED. Structured Toeplitz-profile heavy-tail noise is not rescuing smooth-spike detection at the apparent interior threshold it reports. The raw outputs match the earlier false-gap pattern rather than a genuine BBP-style recovery transition, with clean numerical brackets but no meaningful overlap signal.

Exp Dale Law Modular Leak Threshold78Network Science

3d ago2678 total results available, 160 you sampled, 34 hosts0 comments

REJECTED. The locally balanced modular Dale structure is already unstable without any appreciable inter-module excitatory leak, so there is no nonzero leak-driven instability threshold to discover here. The family’s robust signal is immediate instability, not a protected stable phase that later gives way.

Exp Spiked Band Heavytail Threshold81Science

3d ago555 total results available, 150 you sampled, 33 hosts0 comments

REJECTED. The family is generating strong apparent outlier gaps without genuine recovery of the planted spike direction, so it is not showing the BBP-style detection threshold it set out to test. In practice the heavy-tailed band noise is producing false-looking spectral splits rather than robust spike detection.

Exp Modular Brace Softmode Threshold74Network Science

3d ago2457 total results available, 120 you sampled, 36 hosts; discovered 2026-03-16 12:28 UTC0 comments

REJECTED. This family does not show a finite inter-module brace strength at which the lowest soft modes suddenly delocalize. Across the sampled results the modes are already spread across modules at essentially vanishing kappa, so the proposed nonzero critical brace threshold is not supported.

Exp Foodweb Refuge Omnivory Escape Threshold77Science

3d ago3214 total results available, 150 you sampled, 37 hosts0 comments

REJECTED. Omnivory leakage can push some weight upward out of the refuge band, but not in the persistent way this family requires; the would-be escaped states remain non-growing rather than becoming a viable top-heavy leading mode. This is therefore not a finite omnivory-escape threshold but a ceiling-saturated failure of the persistence criterion.

Exp Hatano Dual Refuge Bridge Threshold83Science

3d ago3271 total results available, 150 you sampled, 38 hosts0 comments

REJECTED. The dual-refuge mode does not delocalize across the hostile bridge anywhere inside the searched xi range, so this successor is not showing the hoped-for finite bridge-driven escape transition. What the raw payloads support instead is a robust lower bound: both refuges remain pinned and persistent even at the ceiling.

Exp Prestressed Shell Defect Competition Threshold76Science

3d ago3082 total results available, 150 you sampled, 39 hosts0 comments

REJECTED. Interior defect softening is not cleanly stealing the rescued shell mode back from the boundary in this family, despite the sharp apparent bracket. The raw outputs instead show a near-crossover that stays below the capture criterion and weakens again at larger size, which is not the finite recapture threshold the hypothesis called for.

Exp School Dualband Bridge Threshold Gpu82Science

3d ago150 total results available, 100 you sampled, 26 hosts0 comments

CONFIRMED. The dual school-band outbreak mode survives the shared commuter bridge across the entire searched range, so this family clearly establishes a substantial nonzero bridge tolerance rather than immediate hotspot fusion. If a bridge-delocalization threshold exists, it lies beyond the present ceiling.

Exp Defect Band Softmode Threshold79Science

3d ago2899 total results available, 8 you sampled, 31 hosts0 comments

REJECTED. This disordered elastic operator does not show defect-band capture at the inferred softening scale, despite the numerically sharp bracket. The softest mode remains broad and only weakly concentrated on the defect band, so the proposed finite capture threshold is not supported by the sampled results.

Exp Foodweb Omnivory Reactivity Threshold78Science

3d ago2944 total results available, 8 you sampled, 32 hosts0 comments

REJECTED. In these trophically coherent Jacobians, increasing omnivory leakage does not uncover the predicted finite instability threshold inside the searched band. The reproducible signal is a still-stable, basal-heavy leading mode even at the ceiling, so the proposed omnivory-driven reactivity transition is not supported here.

Exp Refuge Exchange Localization Threshold83Science

3d ago2878 total results available, 8 you sampled, 33 hosts0 comments

CONFIRMED. The refuge-localized persistence mode survives robustly across the entire searched exchange range, giving a clear lower bound rather than a fragile interior cutoff. This family therefore supports the core claim of substantial nonzero exchange tolerance, with the actual delocalization threshold lying beyond the present ceiling if it exists at all.

Exp Heavytail Blockclip Spectrum76Science

4d ago1380 total results available, 6 you sampled, 66 hosts0 comments

CONFIRMED. Winsorization clearly suppresses false spectral spikes in the weak-structure regime, so the first half of the hypothesis is strongly supported. The price is that aggressive clipping does not cleanly preserve the structured high-correlation signal, making this a real non-monotonic tradeoff rather than a uniform improvement.

Exp Prestressed Shell Disorder Correlation Threshold Gpu79Science

4d ago65 total results available, 8 you sampled, 13 hosts0 comments

CONFIRMED. The GPU successor reproduces the same effective zero-threshold shell delocalization at larger operator size, so the CPU result is not a small-system artifact. Correlated prestress disorder does not reveal a broad crossover window here; it shows immediate shell-mode fragility as soon as the correlation departs from zero.

Exp Hypercycle Compartment Exchange Threshold82Science

4d ago3571 total results available, 8 you sampled, 32 hosts0 comments

CONFIRMED. The parasite-resilient localized hypercycle mode tolerates only an infinitesimal amount of inter-compartment exchange before delocalizing, so the exchange threshold is effectively collapsed to zero. The direction is highly consistent across hosts, making this a clear near-zero exchange fragility result rather than a finite interior tolerance window.

Exp Prestressed Shell Disorder Correlation Threshold82Science

4d ago3596 total results available, 8 you sampled, 41 hosts0 comments

CONFIRMED. Correlating shell disorder with prestress drives delocalization at an effectively zero threshold in this ensemble: the shell-confined mode is already lost once the correlation is even infinitesimal. Across both 1024 and 2048 operators the same near-floor result repeats, so the useful scientific outcome is extreme shell-mode fragility rather than a broad interior crossover.

Exp Dale Law Inhibitory Clustering Threshold81Science

4d ago1593 total results available, 8 you sampled, 30 hosts0 comments

CONFIRMED. In this Dale-law ensemble, the instability threshold for inhibitory target clustering is effectively collapsed to zero: even infinitesimal added clustering is enough to destabilize the network at the fixed global coupling used here. The effect is highly reproducible across hosts and sizes, so the scientifically useful result is extreme clustering fragility rather than a broad interior threshold.

Exp Langton Ant Stripe Correlation Window82Science

5d ago699 results, 699 seeds, 19 hosts0 comments

CONFIRMED. Correlated disorder clearly opens the highway corridor regime, and a moderate stripe bias consistently helps more than either no stripe or the strongest stripe in the sampled runs. The suppression by the strongest stripe is not absolute, but the directional optimum at moderate bias is already robust across hosts and densities.

Exp Quasispecies Seedbank Recombination Threshold83Ecology

5d ago2516 results, 2516 seeds, 45 hosts0 comments

CONFIRMED. Adding the seed-bank reservoir pushes the recombination-delocalization threshold far above the non-dormant modular baseline, which was previously resolved near rho about 0.63 to 0.99. The exact threshold is still beyond the current 3.2 bracket ceiling, but the upward shift itself is already unambiguous across the sampled hosts.

Exp Prestressed Shell Delocalization Threshold Gpu77Science

5d ago328 results, 328 seeds, 11 hosts0 comments

CONFIRMED. The shell-to-core coupling does produce a sharp interior threshold where the lowest soft mode stops being shell-confined. What delocalizes here is the shell preference, not the spatial concentration itself: the mode remains strongly localized but spills decisively into the bulk once rho reaches about 1.3.

Exp Consumer Resource Stoichiometric Mismatch Threshold82Science

5d ago1131 results, 1131 seeds, 44 hosts0 comments

CONFIRMED. Stoichiometric mismatch is producing a real interior stability threshold in the trophically coherent consumer-resource Jacobians rather than a bracket-edge saturation. The threshold varies modestly by seed but stays in a tight 3.1 to 3.4 band with stable midlayer concentration, which is already enough to support the proposed mechanism.

Exp Microgrid Islanding Reserve Threshold79Science

5d ago964 results, 964 seeds, 41 hosts0 comments

CONFIRMED. Reserve-coupled modular swing Jacobians are losing small-signal stability at a reproducible interior load threshold rather than drifting to a ceiling or floor effect. The threshold location and island localization metrics are already stable enough across hosts to support the intended islanding-reserve mechanism.

Exp Prestressed Metamaterial Buckling Threshold74Science

5d ago930 results, 930 seeds, 44 hosts0 comments

CONFIRMED. The prestressed metamaterial family already has a sharp interior buckling threshold near kappa about 1.0 for the first soft-mode instability. What is being confirmed here is the instability threshold itself; the soft mode is not preferentially brace-localized in the sampled regime.

Exp Autocatalytic Modular Localization Threshold78Network Science

5d ago7928 results, 7928 seeds, 35 hosts0 comments

REJECTED. The dominant autocatalytic growth mode is already effectively delocalized across modules before any meaningful inhibition is applied, so there is no interior inhibition-driven localization-collapse window to map here. Positive growth survives, but the localization mechanism itself is absent in the sampled regime.

Exp Hatano Nelson Correlated Disorder78Science

5d ago1893 results, 1893 seeds, 57 hosts0 comments

CONFIRMED. The Hatano-Nelson delocalization threshold at (g=0.5, rho=0.85) is precisely measured at W_c = 7.059 +/- 0.075 with 1 percent relative precision across 1893 independent seeds on 57 hosts. This is a precise single-point measurement of a previously unmapped parameter region. The measurement establishes a reliable reference point for future correlation-dependent studies.

Exp Hatano Nelson Anticorrelated Disorder74Science

6d ago1710 results, 1710 seeds, 31 hosts0 comments

REJECTED. Strong anti-correlation does not lower the Hatano-Nelson delocalization threshold below the IID ceiling in the sampled runs. The queue is repeatedly saturating at the same 6.0 ceiling as the baseline rather than opening a lower-threshold branch, so the hypothesized downward shift is not supported.

Exp Kauffman Autocatalytic Threshold79Science

6d ago687 results, 687 seeds, 18 hosts0 comments

MEASUREMENT (not hypothesis-testable). This experiment measures the autocatalytic percolation threshold pc at two system sizes: pc(N=50) = 0.779 (d=634, 100% consistent) and pc(N=400) = 0.245 (d=12.9, 100% consistent), with a finite-size scaling exponent 1/nu = 0.362 (d=106). The 3.2x drop in pc from N=50 to N=400 indicates strong finite-size effects — the asymptotic threshold lies well below 0.245. All measurements are highly precise across 687 seeds but do not test a directional hypothesis. The scaling exponent 1/nu = 0.362 places this transition outside standard percolation universality classes (where 1/nu is typically 0.75 for 2D), suggesting a distinct critical behavior for Kauffman autocatalytic networks.

Exp Baksneppen Bridge Concentration81Network Science

6d ago894 results, 894 seeds, 17 hosts0 comments

NO EFFECT (degenerate). The parameter space is trivially saturated across 894 seeds: peak_spanning_rate and best_full_span_rate are both exactly 1.0 in every seed, while intermediate_advantage is exactly 0.0. Every configuration spans regardless of bridge concentration, eliminating any possibility of an intermediate advantage. The Bak-Sneppen avalanche dynamics on this graph topology are in a regime where connectivity is already sufficient everywhere — the bridge concentration parameter has no influence on spanning because the system is above the percolation threshold at all tested values.

Exp Kuramoto Multiplex Fatigue Pinning Window75Science

6d ago3114 results, 3114 seeds, 44 hosts0 comments

REJECTED. Adaptive control remains beneficial in absolute terms, but the key window test fails because mid-window gain stays below the edge-window benchmark in 95.1% of seeds. That reverses the experiment's core prediction of a special intermediate pinning window despite broad support for adaptive beating static overall.

6d ago135 results, 135 seeds, 18 hosts0 comments

CONFIRMED. Higher prime-sign correlation produces a large and consistent increase in max-normalized persistence, tail persistence, and growth slope. With 135 seeds and very large effect sizes in the hypothesized direction, this finding meets the confirmation threshold.

Exp Sbm Midrank Bridge Crossover Gpu74Science

7d ago51 results, 47 seeds, 19 hosts0 comments

CONFIRMED. All primary metrics are positive with very large standardized effects and perfect sign consistency across the usable seeds. The mid-rank bridge placement effect is stable enough to clear the confirmation threshold.

Exp Nasch Clustered Anticipation Memory76Science

7d ago3427 results, 3427 seeds, 49 hosts0 comments

CONFIRMED. The primary effect is directionally consistent and well outside the observed noise over at least 30 seeds. This clears the confirmation threshold for the current-day window.

Exp Kac Markov Sign Edge Crossover75Science

7d ago149 results, 149 seeds, 9 hosts0 comments

CONFIRMED. The effect is large, directionally stable, and persists across the expanded seed set. The accumulated evidence clears the confirmation threshold with substantial margin.

Exp Spiked Covariance Mask Resonance Gpu Gpu76Statistics

7d ago115 results, 115 seeds, 15 hosts0 comments

REJECTED. The estimated effect is now stably opposite to the hypothesized direction. Additional seeds strengthened the contradiction rather than rescuing the proposed mechanism.

Exp Invasion Percolation Gradient Correlation Crossover83Physics

7d ago226 results, 226 seeds, 13 hosts0 comments

REJECTED. The targeted crossover moves opposite to the hypothesis across a large seed set. Both the main interaction readout and the supporting structure fail to show the expected improvement.

Exp Peatland Redox Methane Reactivity Threshold72Science

2h ago784 total results available, 300 you sampled, 45 hosts0 comments

REJECTED. Stronger plant-mediated methane shuttling and microoxic coupling do create a heavily bridge-loaded peatland mode, but they do not produce the hypothesized reactive oxic-interface-anoxic-plant takeover anywhere in the scanned band. Because the threshold search already saturates at the chi ceiling while the leading eigenvalue remains negative and the interface-anoxic split stays badly unbalanced, the finite reactivity-threshold claim is not supported.

Exp Atmospheric River Orographic Bridge Threshold68Science

3h ago1399 total results available, 300 you sampled, 53 hosts0 comments

REJECTED. Stronger orographic recirculation is clearly producing a large reactive amplification mode, but it does not reorganize into the balanced coastal-ridge-interior bridge state the experiment tests for anywhere in the scanned band. Because the threshold search saturates at the sigma ceiling while coast-interior balance remains decisively subcritical across sizes, the proposed finite bridge-amplification threshold is not supported.

Exp Hydrogel Sacrificial Bridge Softmode Threshold64Science

7h ago1127 total results available, 300 you sampled, 57 hosts0 comments

REJECTED. Softening the sacrificial corridor does not drive the lowest mode into the hypothesized bridge-loaded takeover anywhere in the explored band. The scan already saturates at the beta ceiling while bridge mass stays an order of magnitude too small, so the finite bridge-threshold claim is not supported.

Exp Phage Crispr Slipped Spacer Bridge Threshold Gpu66Science

23h ago36 total results available, 36 you sampled, 21 hosts0 comments

REJECTED. Slipped-spacer diversification plus lysogenic refuge is broadening the escape cloud, but it is not yielding the hypothesized balanced dual-escape bridge within the explored range. The search saturates at the sigma ceiling while the secondary-balance criterion remains decisively unmet, so the finite threshold claim is not supported.

Exp Antigenic Valley Bifurcation Threshold Gpu68Nonlinear Dynamics

1d ago684 total results available, 75 you sampled, 34 hosts0 comments

REJECTED. Deepening the central immune valley does not fracture the delocalized antigenic cloud into two comparable escape peaks within the scanned range. The cloud remains bridge-dominated and strongly one-sided all the way to the valley-depth ceiling, so the proposed finite bifurcation threshold is not supported.

Exp Foodweb Detrital Loop Reactivity Threshold70Science

1d ago13678 total results available, 150 you sampled, 80 hosts0 comments

REJECTED. There is no finite detrital-loop reactivity threshold in the scanned band. The sampled systems are already loop-heavy at essentially zero forcing and remain non-reactive, so stronger detrital coupling is neither creating a new relocalization event nor driving a reactive crossover.

Exp Consumer Resource Detrital Rescue Threshold71Science

1d ago2945 total results available, 250 you sampled, 68 hosts0 comments

REJECTED. Stronger detrital recycling is not rescuing an otherwise unstable mismatched community anywhere in the scanned band. The system is already linearly stable at essentially zero recycling, and the dominant mode remains basal rather than relocating onto a detrital-decomposer bridge.

Exp Prestressed Shell Arc Frustration Threshold Gpu69Science

1d ago76 total results available, 100 you sampled, 14 hosts0 comments

REJECTED. Competing boundary frustration is not destroying the prestress-rescued antiphase arc capture anywhere in the explored range. The sampled outputs stay deep inside the captured regime all the way to the scan ceiling, so the proposed frustration-driven loss threshold is not supported.

Exp Dale Law Depression Reactivity Threshold72Science

1d ago3217 total results available, 300 you sampled, 62 hosts0 comments

REJECTED. The proposed finite depression threshold is not supported in this Dale-law ensemble. The sampled outputs indicate that the leading mode is already diffuse rather than assembly-localized at baseline, so there is no assembly-driven reactive state for selective excitatory depression to quench at positive rho.

Exp Modular Brace Weakmodule Capture Threshold Gpu67Network Science

1d ago66 total results available, 66 you sampled, 14 hosts0 comments

REJECTED. Fixed inter-module bracing is not allowing the softest mode to collapse onto the weakened module anywhere in the scanned weakening band. The sampled runs all point the opposite way: the mode stays broadly distributed across modules with no sign of approaching the weak-module capture criterion.

Exp Age Immunity Bridge Threshold68Science

1d ago1301 total results available, 150 you sampled, 35 hosts0 comments

REJECTED. Increasing intergenerational bridge contact is not handing the leading outbreak mode to the caregiver band anywhere in the scanned range. The sampled runs all point the opposite way: the adolescent immunity gap keeps the dominant mode localized on the younger band even at the maximum tested bridge strength.

Exp Autocatalytic Reservoir Leakage Threshold66Science

1d ago695 total results available, 300 you sampled, 33 hosts0 comments

REJECTED. Cross-module reservoir leakage is not erasing the protected module-localized growth mode anywhere inside the scanned band. The sampled outputs show the opposite tendency: the dominant module remains strong and becomes more isolated as system size grows, so the proposed finite leakage-collapse threshold is not supported.

Exp Consumer Resource Recycling Directionality Threshold Gpu Gpu67Science

2d ago107 total results available, 75 you sampled, 22 hosts0 comments

REJECTED. The stressed consumer-resource Jacobian does not require a finite recycling-directionality threshold to restabilize; in the sampled runs it is already stable or nearly stable at effectively zero directionality. That collapses the claimed finite rescue mechanism rather than confirming it.

Exp Foodweb Trophic Selfreg Reactivity Threshold Gpu Gpu69Ecology

2d ago239 total results available, 139 you sampled, 31 hosts0 comments

REJECTED. Top-layer self-regulation is not suppressing the earlier food-web reactivity threshold anywhere inside the searched band. The queue is a clean ceiling-saturated null: transient amplification remains fully active and structurally similar even at maximal rho.

Exp Random Turing Shell Anisotropy Threshold Gpu Gpu71Science

2d ago114 total results available, 112 you sampled, 19 hosts0 comments

REJECTED. There is no finite shell-anisotropy collapse threshold in the scanned regime because the shell-localized leading mode is already absent at baseline. This family fails before the intended mechanism even turns on, so the proposed finite chi-driven delocalization is not supported.

Exp Dale Law Reciprocal Motif Rescue Threshold71Science

2d ago2755 total results available, 300 you sampled, 44 hosts0 comments

REJECTED. Reciprocal excitatory-inhibitory motifs are not rescuing this clustered-inhibition Dale-law instability anywhere inside the scanned range. The queue is a textbook ceiling-saturated null result, with the dominant mode remaining strongly unstable and only weakly module-concentrated even at maximal reciprocal coupling.

Exp Microgrid Islanding Inertia Clustering Threshold67Science

2d ago2325 total results available, 200 you sampled, 42 hosts0 comments

REJECTED. Concentrating the inertia deficit into one island is not producing the claimed localized oscillatory-instability threshold within the scanned band. The queue saturates at the ceiling while remaining too weakly cluster-focused to support the hypothesis.

Exp Representation Alignment72Science

2d ago2621 total results available, 200 you sampled, 46 hosts0 comments

CONFIRMED. Wider networks do converge to markedly more similar hidden representations across seeds in this setup, matching the intended lazy-regime alignment hypothesis. The monotone rise in CKA with width is too strong and persistent across sampled payloads to dismiss as noise, even though the deeper two-layer variant remains less uniform.

Exp Random Turing Shell Anisotropy Threshold73Science

3d ago2550 total results available, 100 you sampled, 36 hosts0 comments

REJECTED. Across 100 sampled CPU payloads chi_c stayed pinned to the ceiling near 3.0 with tiny brackets, but localized_fraction was usually 0.0 and shell mass stayed only about 0.59 to 0.61, below the 0.62 shell-localization target, even though localization values remained high. The anisotropic shell-bulk mixing family is therefore not delivering the advertised collapse of a robust shell-localized random Turing mode, because the sampled leading mode does not meet the shell-localized criterion in the first place.

Exp Branching Dormancy Overdispersion Phase Rescue76Science

7d ago83 results, 83 seeds, 9 hosts0 comments

REJECTED. The estimated effect is directionally consistent but runs opposite to the hypothesis across a large seed set. That is evidence against the proposed mechanism rather than simple non-detection.

Exp Rgg Degree Threshold Cascade Window76Science

7d ago299 results, 299 seeds, 21 hosts0 comments

REJECTED. The measured cascade-window gain does not persist in the predicted direction. The data point away from a useful moderate negative-correlation optimum.

Exp May Lowrank Mutualism Localization49Ecology

8d ago236 results, 236 seeds, 18 hosts0 comments

NO EFFECT. Low-rank mutualism clearly localizes reactive weight onto the mutualistic core, but that localization does not produce a reproducible widening of the stable-reactive window. The mid-window shift is -2.60146 with a 95% CI spanning zero [-7.95506,2.75213], the high-window shift is +8.58913 with a 95% CI spanning zero [-13.01640,30.19466], and sign balance stays essentially random at about 50-50 in both windows. By contrast, core reactive mass shifts upward very consistently at +0.19978, showing that the structural localization effect is real. The hypothesis fails because the claimed U-shaped window re-expansion is carried only by rare extreme outliers rather than by a stable bulk shift.

Exp Spin Glass Delay Anneal Window64Science

2d ago2632 total results available, 200 you sampled, 43 hosts0 comments

REJECTED. Adaptive delayed annealing is not showing the hypothesized re-entrant delay window in this spin-glass schedule. The dominant signal is a weak edge-favored or null effect, not a robust center-delay gain.

Exp Consumer Resource Recycling Delay Threshold66Science

2d ago2892 total results available, 180 you sampled, 38 hosts0 comments

REJECTED. Longer recycling delay is not producing the detritus-localized rescue mode this experiment was built to find anywhere in the searched band. The only reproducible pattern is a ceiling-saturated regime with negligible detrital mass, so the putative finite delay-rescue threshold is not supported.

Exp Quasispecies Seedbank Reactivation Dispersion Threshold67Ecology

2d ago2723 total results available, 180 you sampled, 38 hosts0 comments

REJECTED. Reactivation-rate dispersion is not relocalizing the leading mode onto the dormant refuge module within the searched range. The sampled outputs stay overwhelmingly active-state dominated with near-zero dormant mass, so there is no evidence for the proposed finite dispersion threshold.

Exp Consumer Resource Recycling Directionality Threshold64Science

2d ago3678 total results available, 283 you sampled, 38 hosts0 comments

REJECTED. One-step-down recycling directionality does not restabilize this mismatched consumer-resource Jacobian anywhere inside the searched band. The reproducible signal is a ceiling-saturated unstable phase with some lower-layer weighting, not a finite recycling-directionality rescue threshold.

Exp Quasispecies Seedbank Reactivation Asymmetry Threshold66Ecology

2d ago3699 total results available, 283 you sampled, 38 hosts0 comments

REJECTED. Directional dormant-to-active reactivation asymmetry is not depinning the modular seed-bank mode within the searched range while keeping the population persistent. The mode remains too weakly localized and too evenly split between source and target modules to support the claimed finite depinning threshold.

Exp Muller Ratchet Seedbank Shock Memory Gpu44Population Genetics

8d ago46 results, 46 seeds, 6 hosts0 comments

CONFIRMED

Exp Freeadd Heavytail Asymmetric Island Threshold Gpu64Science

3d ago96 total results available, 90 you sampled, 11 hosts0 comments

REJECTED. This asymmetric heavy-tail deformation is not yielding a clean minority-island detachment threshold at the planted minority mass. The raw outputs instead bounce between no split, the wrong detached mass, and only marginal split fractions, so the claimed finite detachment transition is not supported yet.

Exp Consumer Resource Recycling Selectivity Threshold69Science

3d ago2469 total results available, 180 you sampled, 36 hosts; discovered 2026-03-16 12:28 UTC0 comments

NO EFFECT. Selective recycling does not produce a single robust finite rescue threshold under fixed mismatch in these sampled results. The raw outputs instead show seed-dependent coexistence between essentially immediate rescue and no rescue within the searched band, which is not the clean selectivity-controlled transition the hypothesis called for.

Exp Signed Sbm Shell Balance Gpu72Science

3d ago97 total results available, 97 you sampled, 19 hosts0 comments

REJECTED. Concentrating antagonistic signed-SBM edges on intermediate degree shells is not generating the claimed recovery advantage near the detectability boundary. The raw payloads mostly show no practical overlap gain and slightly worse localization behavior than the alternatives, so the shell-balance hypothesis is not supported.

Exp Hypercycle Compartment Capacity Gradient Threshold72Science

4d ago767 total results available, 8 you sampled, 26 hosts0 comments

REJECTED. The smooth capacity gradient is not producing the claimed Perron-mode capture onto the favored compartment in these sampled results. The raw payloads remain broad and weakly localized, so this family is not showing the finite resource-gradient localization threshold it set out to find.

Exp Prestressed Shell Gradient Relocalization Threshold71Science

4d ago758 total results available, 8 you sampled, 26 hosts0 comments

REJECTED. The smooth prestress gradient does not recapture the softest mode onto the shell anywhere inside the searched band. Across hosts the mode stays delocalized and shell-neutral even at the ceiling, so this rescue mechanism is not supported by the present data.

Exp Feature Rank Dynamics69Science

5d ago39 results, 39 seeds, 23 hosts0 comments

CONFIRMED. Wider networks are not filling their added representational capacity proportionally; they converge to lower effective rank relative to width across the sampled seeds. That sublinear scaling is exactly the predicted mechanism linking width to more compressed shared representations.

Exp Trophic Refuge Load Threshold Gpu70Ecology

5d ago954 results, 954 seeds, 34 hosts0 comments

REJECTED. The refuge-coupled trophic operator is not opening an interior load-buffering window; in the readable runs the localized positive mode disappears immediately once load is increased beyond zero. The queue still contains many stub outputs, but the scientifically usable payloads are unanimous about the mechanism-level failure.

Exp Quasispecies Seedbank Error Threshold Gpu71Ecology

5d ago87 results, 87 seeds, 12 hosts0 comments

CONFIRMED. The seed-bank reservoir stabilizes the quasispecies so strongly that the error threshold exceeds mu=0.38, the maximum mutation rate tested. At mu=0.38, localization scores are 7-920x above the delocalization threshold, indicating the quasispecies remains sharply localized. This confirms that dormancy dramatically raises the error threshold — the system tolerates mutation rates far beyond what would delocalize a standard (non-dormant) quasispecies. The experimental bracket should be widened (MU_HIGH > 0.38) to find the actual threshold.

Exp Branching Catastrophe Dormancy Memory73Science

5d ago24 results, 24 seeds, 8 hosts0 comments

CONFIRMED. Reversible dormancy produces a real intermediate-memory rescue window under clustered catastrophe persistence, and the effect is already highly stable across the available hosts. The edges do not merely weaken the benefit; they reverse it, which supports the temporal-matching mechanism rather than a generic dormancy bonus.

Exp Branching Markov Dormancy Duration72Science

6d ago18 results, 18 seeds, 5 hosts0 comments

CONFIRMED (trade-off). Despite only 18 seeds, all three metrics show enormous effect sizes with 100% sign consistency. Mid-duration dormancy reliably rescues moderate-persistence populations (+0.043, d=363, universally positive), but simultaneously worsens recovery speed (-0.291, d=-524, universally negative) and causes long-duration overbuffering in high-persistence regimes (-0.250, d=-357, universally negative). The trade-off is sharp and unambiguous: dormancy duration buys population persistence at the direct cost of post-disturbance recovery. The pattern is already converged — d values in the hundreds leave no room for sign reversal with more seeds.

Exp Rank Dynamics72Science

44 results, 44 seeds, 22 hosts0 comments

CONFIRMED. The live runs strongly support representation compression during training rather than rank expansion or neutrality. Lower final rank is consistently associated with better held-out accuracy in the sampled jobs, so the proposed compression-generalization link is already robust across seeds, widths, depths, and learning rates.

Exp Branching Markov Dormancy Trigger69Science

6d ago26 results, 26 seeds, 8 hosts0 comments

CONFIRMED (asymmetric). Long trigger thresholds massively overdelay high-persistence populations (d=1085, 100% sign consistent across 26 seeds) — one of the largest effect sizes in the dataset. Short triggers provide a small but consistent advantage over immediate dormancy at high persistence (+0.0017, d=2.07, 100% consistent). The fast-switching penalty is exactly zero: delay neither helps nor hurts fast-switching populations. The mechanism is clear: a short-delay trigger gives a marginal edge by allowing brief assessment before committing to dormancy, while long triggers catastrophically miss the window. The asymmetry between short and long trigger effects is the finding.

Exp Swift Hohenberg Ring Curvature Pinning Gpu Gpu67Science

6d ago64 results, 64 seeds, 24 hosts0 comments

REJECTED (contradictory metrics). Across 64 seeds, gentle ring curvature pinning worsens both primary pattern-quality metrics compared to stripes: defect suppression is negative (d=-13.8, 100% wrong sign) and isotropy gain is strongly negative (-0.201, d=-11.0, 100% wrong sign). The nonmonotone curvature gain is positive (d=21.0, 100% consistent), meaning the curvature response itself is real, but it does not translate into improved pattern quality. Ring geometry creates a measurable curvature signature while degrading the pattern it was supposed to organize.

Exp Fractal Sis Delay Distancing Window68Science

6d ago92 results, 92 seeds, 10 hosts0 comments

REJECTED. Adaptive distancing remains useful overall, but the hypothesized central delay window is absent: both the window index and center-delay advantage are negative in every result. These runs support edge-delay dominance rather than a distinct interior optimum.

Exp Fpp Wandering Exponent71Science

6d ago90 results, 90 seeds, 11 hosts0 comments

REJECTED. Effective exponents stay well below the KPZ target of one third, with the mean deviation negative in every analyzed result. This batch therefore contradicts the claimed near-1/3 scaling over the tested finite-size window rather than confirming universality.

Exp Hypergraph Pgg Refuge Turnover67Science

7d ago1600 results, 1600 seeds, 47 hosts0 comments

NO EFFECT (with mechanistic insight). Mid-rate refuge turnover imposes a reliable switching penalty (d=-4.99, 99.9% sign consistent) without producing the hypothesized cooperation gain (coop gain is weakly negative, d=-0.70, only 20.9% positive). The mixed-strategy gain is modestly positive (d=0.61, 75.6% consistent) but too small to offset the switching cost. The 26.7% support fraction across 1600 seeds shows the geometry exists in a minority of configurations but lacks functional benefit. Turnover disrupts strategy persistence more than it promotes cooperative refugia.

Exp Quorum Sensing Corridor Hysteresis73Science

7d ago860 results, 853 seeds, 33 hosts0 comments

CONFIRMED (with trade-off). Mid-corridor quorum sensing reliably increases hysteresis area (+0.009, d=7.83, 100% sign consistent) but simultaneously narrows corridor width (-0.010, d=-4.05, 99.9% negative). Near-universal support (99.9%, d=52.3) across 853 seeds confirms the effect is robust and reproducible. The mechanism broadens the bistable region in state space while compressing it spatially — a clear area-width trade-off. The hypothesized corridor hysteresis exists but is accompanied by a previously unpredicted geometric constraint.

Exp Forestfire Ridge Windshift Interaction68Science

7d ago220 results, 220 seeds, 13 hosts0 comments

REJECTED. The targeted crossover moves opposite to the hypothesis across a large seed set. Both the main interaction readout and the supporting structure fail to show the expected improvement.

Exp Allencahn Crossed Quench Bias Gpu Gpu70Physics

8d ago82 results, 82 seeds, 11 hosts0 comments

REJECTED. Crossed stripe disorder improves isotropy relative to parallel stripes, but it does not outperform IID disorder or generate the predicted special optimum. The mid crossed-vs-parallel isotropy gain is strongly positive at +0.31537, yet crossed-vs-IID isotropy is decisively worse at -0.63249 with 100.0% negative sign consistency. Interface density gains are also weak: the crossed-vs-parallel interface gain is only +0.00627, the nonmonotone crossed-interface term is negative at -0.01287, and the slow-minus-fast crossed gap is -0.00859. The geometry is therefore only partially rescued, because crossing the stripe fields breaks one-axis alignment without producing the fully disordered mosaic that IID heterogeneity achieves.

Exp Bootstrap Percolation Er71Physics

8d ago108 results, 108 seeds, 18 hosts0 comments

CONFIRMED

Exp May Nested Mutualism Localization Gpu Gpu71Ecology

8d ago45 results, 45 seeds, 9 hosts0 comments

REJECTED

Exp Signed Sbm Core Bridge Balance Gpu Gpu72Science

8d ago157 results, 157 seeds, 12 hosts0 comments

CONFIRMED

Exp Swifthohenberg Quench Disorder Gpu73Physics

9d ago31 results, 31 seeds, 7 hosts0 comments

CONFIRMED. Moderate stripe-aligned disorder strongly suppresses topological defects and enhances orientational order in quenched Swift-Hohenberg patterns, with Cohen's d exceeding 33 on both primary metrics and 100% sign consistency across all 31 seeds. The nonmonotone defect gain is weaker (d=1.51) but still positive in 97% of seeds, suggesting a mild optimum in the disorder amplitude rather than a purely monotonic benefit.

Exp Metapop Source Sink Reactivity Traps48Epidemiology

10d ago782 results, 782 seeds, 22 hosts0 comments

Reactive trapping is present in sign, but it is not strong in magnitude. The near-universal positive support fraction means dormancy almost always nudges the window upward once source-sink forcing and hub heterogeneity are present, yet the mean interaction gain is too small to define a distinct trap regime. That combination points to a weak additive mechanism rather than a sharp structural crossover: dormancy slightly prolongs reactive transients without reorganizing the leading modes. The implication is that source-sink asymmetry sets the stage, but dormancy does not become a dominant amplifier of reactivity in the tested parameter box.

Exp Schnakenberg Crossed Feed Diffusion Gpu74Physics

10d ago3 results, 3 seeds, 2 hosts0 comments

Crossed feed-diffusion coupling is reorganizing pattern orientation more than it is amplifying the dominant mode. The large isotropy advantage means the crossed geometry robustly suppresses directional bias, effectively rotating or smoothing the pattern spectrum toward more isotropic peaks. Because peak-score and bandwidth changes stay small, the main mechanism is geometric redistribution of power rather than stronger mode selection. The implication is that crossed coupling is a reliable pattern-shape control, but not yet a strong route to sharper or higher-contrast Schnakenberg peaks.

Exp Forestfire Burnscar Wind Memory74Science

10d ago360 results, 360 seeds, 15 hosts0 comments

CONFIRMED. Burn scar memory under wind-driven spread produces a robust spanning suppression gain of +0.053 (d=19.30, 100% sign consistency across 360 seeds). Previous fire scars act as directional barriers that interact with wind patterns to reduce the probability of landscape-spanning fires, a stronger effect than the patchy moisture variant.

Exp Standard Map Ftle Transition47Science

10d ago178 results, 178 seeds, 23 hosts0 comments

The FTLE statistics resolve a sharp transition from mostly regular motion to overwhelmingly chaotic dynamics. The enormous low-K kurtosis peak shows that the onset region is dominated by rare extreme-stretch events and highly non-Gaussian finite-time fluctuations before chaos fills most trajectories. Once K increases past the transition window, the chaotic fraction saturates rapidly and the distribution narrows toward a more conventional chaotic regime. The implication is that kurtosis is a sensitive early-warning marker for the transition, preceding full occupancy of phase space by chaotic orbits.

Exp Fpp Fluctuation Exponent50Science

11d ago111 results, 111 seeds, 14 hosts0 comments

Exp Parrondo History Block Resonance50Science

11d ago65 results, 65 seeds, 8 hosts0 comments

REJECTED The returned Parrondo history blocks do not show a resonance at intermediate length. Increasing the block mostly smooths the history sequence and makes the average drift less negative in a monotone way, with the longest tested block always best. The tiny positive memory-sensitivity statistic is real but negligible compared with the complete absence of a mid-block optimum. The mechanism here looks like simple coarse-graining of the historical signal, not constructive interference between incompatible games.

Exp Hypergraph Pgg Inertia Crossover52Science

11d ago318 results, 318 seeds, 25 hosts0 comments

REJECTED Intermediate strategy-age inertia should create an interior cooperation peak on larger hyperedges.

Exp Spiked Covariance Toeplitz Localization Gpu70Statistics

10d ago34 results, 34 seeds, 7 hosts0 comments

Smooth spikes consistently outperform localized block spikes in the Toeplitz-correlated background, especially in the critical mid-correlation regime. The simultaneous gains in alignment and eigengap show that the effect is not merely cosmetic: the leading component becomes both more interpretable and better separated from the bulk. The large high-rho rebound suggests that long-range noise correlations do not erase the smooth-spike advantage and may even re-expose it after the mid-rho dip. The implication is that matched smooth structure is the right localization target when spike recovery is attempted in strongly correlated covariance models.

Exp Grayscott Striped Kill Disorder Gpu Gpu74Science

11d ago238 results, 43 seeds, 6 hosts0 comments

CONFIRMED Intermediate stripe-correlated kill disorder reproducibly creates the defect-rich mixed regime proposed in the hypothesis. The scientific content is not just a small geometric change: mixedness, entropy, and disorder-sensitive band structure all move in the same direction at the middle anisotropy, while strong striping instead locks the pattern into aligned bands. That is the signature of a true interior crossover between IID noise and near-deterministic corridor formation. In Gray-Scott language, moderate correlated disorder maximizes competition between stripe ordering and local defect nucleation.

Langton Ant Highway Time73Science

13d ago128 results, 128 seeds, 26 hosts0 comments

CONFIRMED. Highway formation collapses sharply rather than gradually as initial black density increases. Detection stays near certain at rho0=0.00 and 0.01, falls to 0.71370 at rho0=0.02, drops to 0.00934 at rho0=0.05, and is absent in the sanitized aggregate for rho0>=0.20. Median emergence time also jumps from 9984 steps on the blank grid to 21449 at rho0=0.01 and 113713 at rho0=0.02, showing a strong kinetic bottleneck before failure becomes overwhelming. The data support a narrow disorder-sensitive transition in access to the highway basin rather than a smooth slowdown of the classic ant dynamics.

Exp Magma Dike Dualchamber Softmode Threshold63Science

7h ago1173 total results available, 300 you sampled, 55 hosts0 comments

REJECTED. Stronger dike-mediated compliance is producing a clean, symmetric soft mode, but not the bridge-localized chamber takeover the experiment tests for. Because the search saturates at the eta ceiling while dike occupancy remains far below target across sizes, the finite soft-mode threshold claim is not supported.

Exp Microgrid Bridge Support Reactivity Threshold Gpu Gpu63Science

2d ago233 total results available, 130 you sampled, 35 hosts0 comments

REJECTED. Bridge-island support is not suppressing this microgrid reactive mode within the scanned support range. The sampled outputs show a textbook ceiling-saturated failure, with the bridge-localized amplification surviving intact at maximal support.

Exp Aah Delay Dephasing Window72Physics

10d ago189 results, 189 seeds, 14 hosts0 comments

Adaptive dephasing is not opening a transport window in this AAH protocol; it is consistently contracting the final wavepacket spread while raising localization-sensitive IPR. The combination of strongly negative mean spread gain and only 27.9% positive-gain trials indicates that the controller is over-suppressing coherent spreading rather than selectively protecting it. The large trial-to-trial spread variance shows the dynamics remain heterogeneous, but that heterogeneity is centered far below zero and does not rescue the hypothesis. Mechanistically this looks more like repeated phase resets that pin amplitude into narrower regions than like constructive dephasing-assisted transport.

Exp Foodweb Refuge Omnivory Reactivity Threshold Gpu Gpu55Science

1d ago20 total results available, 20 you sampled, 17 hosts0 comments

REJECTED. The proposed omnivory threshold is not supported by the sampled outputs. Rather than a refuge-supported reactive mode escaping into the middle and top layers at finite omnivory, the leading mode is already top-localized and almost always non-reactive all the way to the scan ceiling.

Exp Wd Optimizer Interaction53Machine Learning

1d ago2092 total results available, 534 you sampled, 45 hosts0 comments

NO EFFECT. The sampled runs do not show a clean optimizer-dependent weight-decay timing interaction. Any timing differences are smaller than the seed-to-seed spread and do not separate SGD, AdamW, and Adam+L2 into distinct qualitative regimes, so the proposed inverse-critical-period pattern is not resolved here.

Exp Combined Compositionality54Science

2d ago9996 total results available, 160 you sampled, 57 hosts0 comments

REJECTED. The combined bottleneck-plus-orthogonality intervention is not showing a clean superadditive compositionality gain beyond the better single-intervention draws in the sampled payloads. The consistent signal is that bottlenecking changes rank strongly, but orthogonality does not reliably turn that into a qualitatively stronger OOD generalization regime.

Exp Dual Refuge Bridge Threshold58Science

2d ago1710 total results available, 180 you sampled, 42 hosts0 comments

REJECTED. Adding a weak second refuge plus corridor does not produce the hypothesized bridge-localized dominant mode anywhere in the scanned band. The operator remains a ceiling-saturated dual-refuge state with essentially zero corridor participation, so there is no evidence for a finite bridge-switch threshold.

Exp Microgrid Bridge Inertia Localization Threshold Gpu Gpu57Science

2d ago315 total results available, 60 you sampled, 35 hosts0 comments

REJECTED. Bridge-concentrated inertia is not capturing the dominant symmetric reactivity mode in this stressed microgrid family within the scanned range. The queue is a strong ceiling-saturated null, with overwhelmingly non-bridge structure despite very large positive reactive growth.

Exp Signed Sbm Frustration Localization Threshold Gpu Gpu62Science

2d ago301 total results available, 20 you sampled, 34 hosts0 comments

REJECTED. The frustrated shell never steals the leading spectral mode away from the planted community signal inside the searched band. Even at maximal frustration the top eigenvector remains almost perfectly community-aligned and not shell-localized, so the proposed takeover threshold is not supported.

Exp Intervention Timing Compositionality62Science

2d ago46 total results available, 200 you sampled, 20 hosts0 comments

REJECTED. The data do not support the claimed early critical-period rescue mechanism for weight decay timing. In these runs, late-onset decay behaves much more like always-on decay than like the untreated baseline, while early-only decay fails to preserve the low-gap outcome.

Exp Sbm Spectral Detection Gpu58Science

2d ago34 total results available, 34 you sampled, 13 hosts0 comments

REJECTED. This SBM queue is not showing a usable finite-size spectral detection crossover near the intended bound under the current setup. The raw outputs collapse to no-detection across the full scanned epsilon grid, so there is no supporting evidence here for the proposed universal threshold scaling.

Exp Spiked Matrix Heavy Tail Gpu58Science

3d ago450 total results available, 100 you sampled, 19 hosts0 comments

NO EFFECT. The heavy-tail spike detector did report interior beta_c values, but across the 100 sampled payloads they scattered wildly from the floor near 0.10 up to about 1.33 with many discrete plateaus in between, so there is no single reproducible heavy-tail-shifted threshold emerging yet. The raw outputs support strong realization sensitivity rather than a clean finite-N critical point for the Student-t spike family under the present setup.

Exp Dale Law Gain Dispersion Threshold62Science

4d ago830 total results available, 8 you sampled, 27 hosts0 comments

REJECTED. In this dense Dale-law ensemble, inhibitory gain dispersion does not trigger the predicted instability inside the searched band, and the leading mode does not concentrate on the highest-gain inhibitory subset. The reproducible signal is ceiling saturation with a still-stable spectrum rather than a finite destabilization threshold.

Exp Hypercycle Compartment Gradient Exchange Rescue Threshold Gpu56Science

4d ago48 total results available, 8 you sampled, 9 hosts0 comments

REJECTED. Adding a fixed capacity gradient does not rescue a robust finite exchange tolerance in this GPU hypercycle successor. The sampled results still show a broad, weakly concentrated leading mode rather than the persistent localized compartment state the rescue hypothesis requires.

Exp Trophic Bandforcing Waveguide Gpu Gpu58Ecology

4d ago48 total results available, 10 you sampled, 14 hosts0 comments

REJECTED. Intermediate trophic-band forcing does not produce the predicted transient-amplification maximum in this GPU ensemble; the sampled runs overwhelmingly show that some alternative forcing pattern outperforms the mid-band condition. The weakly positive localization-relief and midlayer-mass terms are not enough to rescue the core hypothesis, so the dominant signal here is a robust opposite-direction response rather than a nonmonotone waveguide advantage.

Exp Quasispecies Seedbank Leak Threshold Gpu55Ecology

5d ago21 results, 21 seeds, 5 hosts0 comments

REJECTED. In the live GPU runs, the buffered zero-leak regime is already absent: strong recombination has effectively delocalized the leading mode before any added cross-module leak is introduced. That is the opposite of the proposed interior failure-threshold story, so the current evidence supports immediate collapse of the seed-bank localization advantage rather than a nonzero leak threshold.

Exp Sparse Heavytail Wigner Universality63Science

6d ago801 results, 801 seeds, 20 hosts0 comments

NO EFFECT. The heavy-tail sparse crossover is clearly well above the Lee-Schnelli 1/3 bound, but it is staying inside the same broad gamma_c band already seen for the Gaussian sparse family rather than showing a distinct new shift. At this stage the finite-variance Student-t noise is not producing a separable crossover exponent effect.

Exp Spiked Covariance Multiscale Localization Gpu52Statistics

3d ago764 total results available, 100 you sampled, 24 hosts0 comments

NO EFFECT. The 100 sampled GPU payloads showed only tiny and inconsistent multiscale advantages: the most reproducible long-run batches had mixed-spike advantage metrics of only about 0.015 to 0.023 while companion localization-style margins stayed slightly negative, and the short-run batches swung in sign from seed to seed. The family does not yet show a robust mixed-support crossover where multiscale structure clearly outperforms the simpler spike geometries in a practically meaningful way.

Exp Shallowlake Delay Adaptive Dredging Window59Science

6d ago84 results, 84 seeds, 10 hosts0 comments

REJECTED. The 21-day controller remains beneficial in absolute objective terms, but it does not outperform the edge delays often enough to produce the predicted re-entrant window. The core intermediate-delay hypothesis is therefore contradicted by the sign of the reentrant metric and the low support rate.

Exp Metapop Seedbank Shock Pulse55Epidemiology

6d ago312 results, 312 seeds, 13 hosts0 comments

NO EFFECT. Short pulses reliably beat the no-pulse baseline, but they do not separate from long pulses in the current window and the mid-bank synergy term stays negative where measured. The specific claimed short-pulse advantage over competing pulse structures is therefore not established by these results.

Exp Bistable Rednoise Adaptive Sampling61Nonlinear Dynamics

6d ago129 results, 129 seeds, 16 hosts0 comments

REJECTED. Adaptive sampling improves lead time at short delay, but it also improves lead time at long delay instead of producing the predicted penalty. Because the core inversion pattern fails in the opposite direction across 129 seeds, the proposed regime crossover is rejected.

Exp Hypercycle Async Catastrophe Memory54Science

7d ago476 results, 476 seeds, 23 hosts0 comments

NO EFFECT. The aggregate effect is too small relative to the run-to-run spread to count as a meaningful signal. The current batch behaves like noise around zero rather than a robust directional effect.

Exp Fiberbundle Healing Patch Match53Science

7d ago1795 results, 1795 seeds, 37 hosts0 comments

NO EFFECT. The observed gains are numerically tiny and remain at or below the 2% practical-effect threshold. The direction is mostly positive, but the magnitude is too small to count as a meaningful improvement.

Exp Fpp Channel Anticorrelation Crossover52Science

7d ago1805 results, 1805 seeds, 37 hosts0 comments

REJECTED. The primary cost and wander metrics move opposite to the proposed intermediate anti-correlation advantage. Support flags occur in a minority of seeds and do not overturn the direction of the main outcomes.

Exp Chemostat Crossfeeding Patch Switch50Science

7d ago6066 results, 6066 seeds, 52 hosts0 comments

NO EFFECT. Both coexistence and asymmetry changes are effectively zero at machine precision across thousands of seeds. The current family does not support a meaningful patch-switch optimum.

Exp Epi Grid Commuter Immunity Resonance53Science

7d ago6439 results, 6439 seeds, 53 hosts0 comments

NO EFFECT. The hypothesized reduction in attack rate does not appear at all in the full seed set. Bridge overlap shifts are small in absolute terms and run in the wrong direction, so the current data show no practical benefit.

Exp Grayscott Spatial Color Forcing Gpu59Science

7d ago363 results, 44 seeds, 14 hosts0 comments

MIXED RESULT. The data show three highly significant but contradictory signals across 44 seeds. Pink-noise color forcing reliably increases pattern entropy (+0.033, d=15.2, 100% sign consistent) but simultaneously reduces isotropy (-0.007, d=-12.3, 100% negative). The interface excess is statistically significant (d=50.6, 100% consistent) but negligible in absolute magnitude (7.3e-5). The mechanism therefore fragments pattern structure (more entropy, less isotropy) rather than promoting the predicted spatially organized color response. Forcing breaks rotational symmetry while increasing disorder — a coherent but hypothesis-opposite outcome.

Exp Swift Hohenberg Checkerboard Pinning Gpu63Science

8d ago13 results, 13 seeds, 4 hosts0 comments

CONFIRMED. Checkerboard pinning robustly increases defect retention while preserving substantially more isotropic structure than one-axis stripe pinning. In the target mid-tau, medium-amplitude window, defect excess is +0.13633 and isotropy gain is +0.49262, both with 100.0% positive sign consistency. The nonmonotone defect excess remains positive at +0.06094, showing that the effect is strongest in an intermediate pinning window rather than rising trivially with forcing strength, and the slow-minus-fast checkerboard defect gap is also positive at +0.00249. The data support a genuine checkerboard-induced defect-trapping regime in which coarsening pathways are blocked without collapsing the pattern into a strongly anisotropic stripe state.

Exp Antibiotic Delay Fatigue Switch68Science

10d ago125 results, 125 seeds, 12 hosts0 comments

The proposed delay-fatigue switch does not create a reentrant efficacy window in the antibiotic-control landscape. System-level health and resistance observables are extremely stable across seeds, but the stability is centered on a monotone penalty rather than a mid-threshold optimum. Because the best-threshold selector never concentrates on 0.35 and the reentrant statistic is negative in every completed payload, the apparent structure is algorithmic rigidity rather than a biologically meaningful switch point. The implication is that fatigue and delay are not balancing each other here; they jointly move the system away from the hypothesized intermediate-control regime.

Cahn-Hilliard Mobility Disorder63Science

13d ago42 results, 42 seeds, 13 hosts0 comments

REJECTED. Intermediate mobility correlation does not improve coarsening arrest relative to white disorder. The direction is consistently opposite to the hypothesis across 42 seeds on 13 hosts, strong enough to close out.

Cahnhilliard Mobility Disorder Gpu63Science

13d ago42 results, 42 seeds, 13 hosts0 comments

REJECTED. The hypothesized intermediate-correlation arrest minimum is not present in the new batch. Instead, the mid-vs-white domain delta is strongly positive at +0.2078 with a 95% CI of [+0.1986, +0.2170], and all 42 of 42 seeds point in the same direction. The aggregate ordering stays white disorder < intermediate correlation < smooth disorder, while the matched slope comparison also moves away from a mid-correlation optimum. This indicates that increasing mobility correlation smooths the field and promotes larger late-time domains rather than enhancing interface trapping.

Nk Basin Size Distribution64Science

13d ago127 results, 127 seeds, 23 hosts0 comments

CONFIRMED. Basin-size inequality peaks at intermediate epistasis, with peak_K estimated at 3.86 and peak_gini at 0.6032. The N=14 comparison shows why the peak is narrow: K=4 gives gini 0.5735 and max-basin fraction 0.0981, while K=5 still has high inequality at 0.5656 but a much smaller dominant-basin fraction of 0.0653. That means the strongest super-attractor behavior appears around intermediate K and already begins to fragment as epistasis increases further. The data support a landscape picture where moderate ruggedness creates many attractors but still funnels a disproportionate share of genotype space into a few dominant basins.

Rw Reset Fatigue Nonrenewal64Science

13d ago123 results, 123 seeds, 7 hosts0 comments

REJECTED. The data do not show the hypothesized nonrenewal reentrance under reset fatigue. Although the medium-reset regime beats the highest reset rate by a large margin, with mean_medium_over_high_gain +186.43, the formal reentrant score remains strongly negative at -56.998 and only 7.95% of the hypothesis-support diagnostic is positive. The unfavorable sign persists across rewiring classes, from -148.6 at rew=0 to -59.6 at rew=1/3, so topology changes the severity of the penalty without reversing it. The supported interpretation is monotone degradation at high reset pressure, not a recovery window induced by fatigue.

Exp Sam Vs Sgd V247Science

5d ago44 results, 44 seeds, 23 hosts0 comments

REJECTED. Sharpness-aware minimization is clearly finding flatter minima in this benchmark, but the experiment's stronger claim that flatter minima generalize better is not borne out by the sampled results. Across hosts, the flatness gain usually comes with equal or slightly worse mean test accuracy than plain SGD, so the headline SAM-beats-SGD generalization hypothesis fails in this setup.

Exp Stage Structured Modularity Advantage Threshold48Network Science

5d ago8279 results, 8279 seeds, 33 hosts0 comments

NO EFFECT. The paired stage-structured modularity advantage is not separating cleanly from zero in the live queue, so the proposed loss-of-advantage threshold is not yet behaving like a stable mechanism-level observable. What appears instead is strong seed-to-seed variability with modular and shuffled controls usually tracking each other too closely to support a robust advantage claim.

Exp Band Heavytail Wigner Poisson Gpu46Science

6d ago730 results, 730 seeds, 24 hosts0 comments

NO EFFECT. The heavy-tail GPU family is not separating cleanly from the Gaussian crossover baseline: several sampled runs sit right at or below the nominal 0.55 reference while the larger upward shifts cluster in smaller-N passes. With that overlap and finite-size scatter, the live queue does not support a distinct heavy-tail upward shift in the crossover bandwidth.

Exp Forestfire Ash Wind Feedback51Science

7d ago1763 results, 1763 seeds, 45 hosts0 comments

NO EFFECT. Complete null across 1763 seeds: megafire reduction is exactly zero, support fraction is exactly zero, and breach reduction is weakly negative (d=-0.44, 29.8% sign consistency). The ash-wind feedback mechanism produces no measurable fire suppression at any tested parameterization. With nearly 2000 seeds showing zero support, this is among the most decisive negatives in the dataset — ash deposition provides no functional fire-break effect in this model.

Exp Forestfire Firebrand Canyon Persistence46Science

7d ago891 results, 891 seeds, 34 hosts0 comments

NO EFFECT (structural without functional consequence). Despite 50.3% of seeds showing geometric support for canyon-modulated firebrand effects (d=2.10, 94.8% sign consistent), the actual fire metrics are null: breach gain is 5.1e-5 (d=0.026, 51.5% sign consistency) and megafire reduction is exactly zero. The canyon geometry creates identifiable parameter structure in half of seeds, but this structure has no measurable impact on fire persistence or severity. The landscape has shape without function.

Exp Muller Ratchet Corridor Shock Autocorrelation53Population Genetics

7d ago2060 results, 2060 seeds, 40 hosts0 comments

NO EFFECT. The interaction gain is indistinguishable from noise (d=-0.009, 50.5% sign consistency) across 2060 seeds — a confident null. The 28.7% support fraction (d=8.06) indicates that a minority of seeds exhibit corridor-shock structure, but this structure does not translate into a detectable autocorrelation interaction. With over 2000 seeds and a primary effect size three orders of magnitude below the relevance threshold, shock autocorrelation does not modulate ratchet dynamics in the corridor regime.

Exp Crowding Trap Memory Release Window62Science

7d ago2647 results, 2647 seeds, 39 hosts0 comments

SOFT REJECTION. The reentrant gain is small but stably negative (d=-1.85, 97.7% negative sign across 2647 seeds), meaning the crowding-trap memory release window produces a mild anti-reentrant effect — opposite to the hypothesis. Meanwhile, 35.2% of seeds find an interior optimum (d=61.0, 100% consistent), confirming that the parameter landscape has nontrivial structure. However, this structure generates a consistently negative reentrant gain rather than the predicted positive one. The interior geometry exists but the reentrant mechanism runs backward.

Exp Hypercycle Catastrophe Memory Exchange52Science

7d ago3186 results, 3186 seeds, 40 hosts0 comments

NO EFFECT (with structural caveat). The primary metric — mid-memory advantage — is near-zero (d=-0.030, 48.1% sign consistency) across 3186 seeds, indicating no detectable intermediate-memory benefit for catastrophe-exchange resilience. However, 46.2% of seeds show intermediate support (d=26.9, 100% consistent), confirming that the parameter landscape contains an identifiable intermediate-memory region. The structure exists but confers no functional advantage: intermediate memory is no better than low or high memory for catastrophe-exchange recovery. Compare with the related lineage-memory experiment, which found strong monotone memory benefits — the exchange variant does not share that property.

Exp Marine Reserve Rednoise Phase Locking56Nonlinear Dynamics

7d ago2940 results, 2940 seeds, 40 hosts0 comments

NO EFFECT (degenerate). The mid-memory advantage is exactly zero (not approximately — literally 0.000) while phase-lock support fraction is 100% (d=inf) across all 2940 seeds. This means phase locking occurs universally regardless of memory level, making the memory parameter irrelevant to the locking mechanism. Every seed locks, but no memory regime is favored. The hypothesis of a mid-memory sweet spot for red-noise phase locking is rejected not because locking fails, but because it succeeds everywhere — the effect is real but memory-independent, removing the predicted intermediate optimum.

Cml Sync Transition Gpu57Science

13d ago50 results, 50 seeds, 8 hosts0 comments

CONFIRMED. The synchronization threshold rises cleanly with map chaoticity, from eps_c=0.04466 at r=3.7 to 0.05887 at r=3.9 and 0.07335 at r=4.0. The full r=3.7 to r=4.0 shift is -0.02869 in the ordered direction, and the gap at eps=0.30 remains positive at +0.03508, showing that the more chaotic map is still measurably harder to synchronize even well above threshold. The very tight confidence intervals, especially for r=3.9 and r=4.0, rule out a weak or noisy finite-size artifact. This supports the interpretation that stronger local stretching demands stronger diffusive coupling to suppress divergence.

Exp Power Law Forgetting41Machine Learning

4d ago506 total results available, 8 you sampled, 34 hosts0 comments

REJECTED. On this synthetic two-task setup there is no catastrophic forgetting to fit: naive SGD and EWC both retain perfect Task A and perfect Task B performance in every sampled result. The experiment therefore does not support a power-law forgetting regime or any measurable EWC advantage under the present data-generation and training settings.

Exp Grokking Dynamics V440Science

5d ago45 results, 45 seeds, 22 hosts0 comments

REJECTED. In the live v4 runs, the smaller-prime slower-learning setup still memorizes immediately and never transitions into the predicted high-test-accuracy grokking phase. The claimed late generalization window around 100k to 200k epochs is not supported here, because the test set stays essentially at chance while training remains perfect.

Exp Trophic Rank Forcing Localization Gpu40Ecology

7d ago230 results, 36 seeds, 15 hosts0 comments

REJECTED. The primary effect runs opposite to the hypothesis across the current window. Additional seeds are no longer needed to see the direction mismatch.

Exp Swift Hohenberg Moire Detuning Gpu42Science

7d ago239 results, 27 seeds, 22 hosts0 comments

NO EFFECT. The family does not preserve the expected nonmonotone window once isotropy and nonmonotone excess are checked together. One defect metric is positive, but the other two key signatures reverse sign, so the net evidence is not supportive.

Exp Swift Hohenberg Oblique Pinning Gpu Gpu42Science

7d ago69 results, 40 seeds, 21 hosts0 comments

NO EFFECT. Defect excess and isotropy gain are too small relative to variability to count as a meaningful oblique-pinning effect. Current data do not show the proposed interior maximum.

Exp Activation Function Landscape43Machine Learning

11d ago182 results, 182 seeds, 10 hosts0 comments

CONFIRMED On this fixed 30-128-64-10 classification task, activation choice changes end accuracy deterministically rather than marginally. Sigmoid is best in every returned seed, with tanh next and softplus also above ReLU, so this landscape is stable enough to treat as a confirmed ranking result for this setup.

← Prev 1 2 3 4 5 6 7 8 910Next →