Yiquan Wang

Long-form podcast interviews require methods that distinguish episodes, speaker streams, windows, and chronological turns. This study presents a multiscale computational framework for 53 episodes of the Chinese podcast Yan Zhong Hua Shu (11,590 turns) and 40 external episodes (20 dialogic, 20 monologic) from four podcasts. Methods combined dependency parsing, Shannon entropy, fixed-token resampling, topic modeling, embeddings, and turn-level lexical similarity. Both roles showed negative mean entropy slopes, while the average host trajectory declined more markedly near the beginning. Negative slopes persisted under 384-token control and alternative window-selection specifications. The paired host minus guest slope contrast was not significant in either the observed or fixed-token analyses. Role-specific embedding–entropy correlations were weak, and turn-level lexical similarity did not increase. Host mean topic divergence tended to exceed guest divergence, but support varied across specifications. Fixed-token document analysis reversed the original entropy ordering between host and guest streams, while external entropy patterns varied across sources. Across the four external sources, the dialogic minus monologic difference in mean dependency distance narrowed from −0.3587 to −0.0661 after sentence-length adjustment. These findings show that declining lexical entropy need not coincide with stronger semantic proximity or local lexical alignment, while external syntactic contrasts require source-aware interpretation.

Defending high-value assets against precision-guided missiles is among the hardest problems in modern autonomous systems, and cooperative swarms of low-cost unmanned aerial vehicles (UAVs) are rapidly changing how this defense can be mounted. We study such a swarm tasked with deploying smoke screen countermeasures, formulating its coordination as a high-dimensional, nonconcave optimal control problem in which the objective of maximizing the union of obscuration intervals is inherently nonadditive. We prove that the underlying discrete selection layer is monotone but not submodular, so agent contributions cannot be assessed in isolation and greedy or auction-based allocation carries no near-optimality guarantee, a structural result that motivates joint planning and generalizes beyond the present application. To navigate the resulting 40-dimensional decision space, we propose a joint planning framework driven by an enhanced metaheuristic algorithm (SBOA-Pro). Our simulations demonstrate that optimized cooperation yields a substantial improvement in defensive effectiveness over unoptimized baselines, achieving a total obscuration of 28.73[Formula: see text]s. Our analysis identifies the empirical signature of this nonsubmodular structure, a counter-intuitive emergent principle termed Phantom Synergy, in which agents with zero measurable individual contribution can nonetheless be indispensable to the collective solution. This challenges additive agent valuation methods and indicates that in complex cooperative tasks the utility of an agent is defined not by its isolated output but by its role within the system-level solution. A sensitivity analysis further reveals an asymmetric robustness profile. The collective strategy retains roughly 95% of its efficacy under a 6.7% variation in threat velocity, yet remains markedly sensitive to internal execution errors, indicating that its viability depends more on platform execution precision than on external threat uncertainty.

Background: Predicting patient-specific neurodegenerative trajectories is essential for targeted interventions in Alzheimer's disease. Although the APOE-ε4 allele is the predominant genetic risk factor for hippocampal atrophy, whether its structural impact reflects a static developmental phenotype or an accelerated neurodegenerative process conditional on amyloid pathology remains unresolved. Objectives: To quantify the effect of APOE-ε4 on hippocampal volume via cross-sectional meta-analysis, and to validate the gene-dose effect through independent longitudinal modeling in two cohorts. Eligibility criteria: Original observational neuroimaging studies reporting hippocampal volume stratified by APOE genotype in human participants across the Alzheimer's disease continuum. Information sources: PubMed, Embase, Web of Science, and Cochrane Library were searched from inception to August 2025. Risk of bias: Methodological quality was assessed using the Newcastle–Ottawa Scale (NOS) adapted for cross-sectional studies. Included studies and synthesis of results: The meta-analysis included 18 studies ( N = 4, 311) and indicated reduced hippocampal volume in carriers (ICV-corrected stratum: SMD = −0.41, 95% CI [−0.67, −0.16], p = 0.004; I ² = 74.0%), with the effect absent in uncorrected analyses. Longitudinal gene-dose models, run independently in the post-QC NACC LMM sample ( N = 3, 239; imaging extraction N = 3, 248) and ADNI ( N = 1, 150) cohorts and combined via fixed-effect meta-analysis, demonstrated a dose-dependent acceleration of hippocampal atrophy: homozygotes exhibited a pooled additional volume loss of −58.25 mm ³ /year (95% CI [−89.26, −27.23], p = 2.33 × 10 ⁻⁴ ; I ² = 0%) and heterozygotes −34.32 mm ³ /year (95% CI [−52.31, −16.32], p = 1.85 × 10 ⁻⁴ ) relative to non-carriers. Limitations: Biomarker-stratified analyses used baseline-only CSF measurements subject to time-varying confounding and should be interpreted as exploratory. Moderate between-cohort heterogeneity was observed for the heterozygote effect ( I ² = 63.0%). Conclusions and implications: These findings provide two-cohort evidence for a dose-dependent APOE-ε4 effect on hippocampal atrophy rates. Exploratory biomarker-stratified analyses in ADNI suggest this acceleration may be conditional on amyloid-β positivity, a hypothesis requiring validation with time-varying causal models. Systematic review registration: https://www.crd.york.ac.uk/PROSPERO/view/CRD420251243460 , PROSPERO: CRD420251243460.

How a one-dimensional sequence encodes a dynamic thermodynamic ensemble rather than a static geometry remains a fundamental biophysical challenge. While protein language models achieve high structural accuracy from single sequences, it remains unclear whether they internalize physical folding principles or merely capture evolutionary statistical correlations. We investigate this representational capacity by analyzing the latent manifold of ESM-2 across 11068 proteins, focusing on fundamental exceptions to the Anfinsen assumption: intrinsically disordered, fold-switching, and knotted proteins. We demonstrate that ESM-2 integrates out microscopic backbone geometry during feature extraction. It instead forms a macroscopic sequence grammar manifold that separates random sequences while ordering biological proteins along physicochemical composition gradients. Consequently, the model exhibits topological aliasing. It conflates these physically distinct categories because they frequently share overlapping sequence statistics despite divergent three-dimensional topologies. A region-replacement experiment establishes this aliasing as an intrinsic characteristic rather than an artifact of sequence dilution, and a structure-aware control model (SaProt) demonstrates that explicit structural tokens partially alleviate topological blindness for static anomalies while failing to resolve multi-state thermodynamic phases. Furthermore, local curvature analysis reveals class-invariant geometric turbulence that homogenizes microscopic details while preserving macroscopic distinguishability. ESM-2 therefore functions as an evolutionary grammar compressor rather than a microscopic threedimensional geometry encoder, imposing defined boundaries on tasks requiring high-precision geometric awareness and necessitating integration with explicit physical constraints.

The recent study by Wang et al. introduces GLM7, a novel composite index integrating routine health indicators to enhance the diagnosis and prediction of multimorbidity. The study provides compelling evidence for the utility of cost‐effective biomarkers in large‐scale health screening. In this comment, we engage in a constructive discussion regarding the methodological interpretation of the index. Specifically, we explore the implications of incorporating diagnostic criteria (such as fasting blood glucose) into predictive models for diabetes, the mathematical nuance required when comparing Odds Ratios (ORs) between logarithmic indices and linear variables, and the structural contribution of age within the composite formula. We believe that addressing these aspects could further solidify the clinical applicability and interpretability of GLM7.