Embodied artificial intelligence is moving from deployable models to persistent agents that learn in the field, acquire skills and migrate across bodies. Governing such a system means governing an individual, not a model, and existing proposals (agent identifiers, activity logs, guardrails) do not survive an agent that keeps rewriting itself. We propose the governable individual: an agent whose competence may change without bound, but whose authority, memory schema, embodiment rights and capability roster can widen only through signed lifecycle transitions that update a public identity commitment. In our tests, neither learned judgement nor behavioural testing was sufficient to carry this on its own; the load-bearing layer must be architectural. We describe the abstraction, a runtime mechanism that realizes it, and the open problems in between.

We propose a mathematical framework for organizational psychology based on a Maximum Entropy model of a group's personalities. The Maximum Entropy model is then decimated to a single ``collective personality''. If the original personality scores are augmented by intelligence and emotional quotients, then a collective intelligence is also mathematically revealed. With simple matrix analyses of the collective personality, one can understand: that weak interpersonal coupling can strongly affect group character; that malleable rather than stubborn personalities control the group's collective personality; that one can mathematically solve for optimal top-down directives to achieve certain group personalities; and that groups can have a personality disorder even if the individuals composing the groups are neurotypical. We hope that this framework provides a useful starting point for future mathematical analyses in organizational psychology related to innate character rather than opinion dynamics or decision making, and note that the analysis can be applied to much more complex Maximum Entropy models than the one proposed here if empirical evidence suggests that the Gaussian model proposed here is overly simplistic.

ssys is a Python package for exact algebraic recasting of supported ODE models into S-system or Generalized Mass Action form. It reads Antimony and SBML models, introduces auxiliary variables through symbolic lifting, and validates transformed systems using symbolic, numerical, and trajectory-based checks. The package provides command-line workflows, notebook generation, and benchmark evidence across curated models and BioModels examples, making classical power-law recasting practical for reproducible systems biology modeling.

With the enormous advances in cerebral imaging techniques, a large amount of data is available for studying the aging and demented brain. In this contribution, we apply the OASIS-3 dataset for identifying small areas of the human gray matter, which have higher- or lower structural connectivity in dementia and aging. As anticipated, we found that finer structures of the hippocampus and the temporal lobe show decreased connectivity in dementia. More surprisingly, the precuneus, the cuneus, and finer structures in the insula show higher connectivity in dementia than in the healthy state.

\textbf{Background:} In Costa Rica, dengue is reported and controlled at the canton level, and outbreaks in one canton are often followed by outbreaks in others. Climate models describe where conditions favor transmission but not how dengue moves \emph{between} places, the directional, between-place spread that shapes where an outbreak travels next. \noindent \textbf{Methods:} From weekly case counts for all 81 cantons (1993--2012; \num{246524} cases) we reconstructed a canton-to-canton spread map using the roughly three-week dengue generation interval, removed the shared seasonal and climatic signal so that only direction-specific spread remained, and summarized it by the receiving and source cantons, an amplification factor, and a directionality index, tracked over five-year windows. \noindent \textbf{Results:} Climate-adjusted spread is strongly directional and concentrates in the lowland Caribbean and Pacific cantons (Limón, Matina, Guácimo, Garabito, Orotina). A local outbreak is amplified about three- to fourfold across the network even though overall transmission is not growing. This amplification was greatest during the emergence phase of the 1990s and declined markedly as annual reported cases increased, while the \emph{direction} of spread remained fixed; the decline persists after controlling for the broadening of surveillance coverage. \noindent \textbf{Conclusions:} Routine surveillance alone can map which cantons tend to experience dengue and the pathways through which it appears to spread, providing a potential input for prioritizing surveillance and vector control, particularly when a serotype or the disease itself is newly establishing. As a historical description of average behavior over multi-year windows, it is a planning input whose prospective value remains to be tested.

Statistical learning is essential for individuals to discover structure in the sensory environment, especially during communication via speech or music. Individual differences in statistical learning abilities have been proposed to account for differences in various cognitive functions and development, including developmental disorders such as dyslexia. In this study, we used a Hierarchical Bayesian Statistical Learning (HBSL) model to model individual learning trajectories as recorded using electroencephalography (EEG) while adults with and without dyslexia listened to structured tone sequences. Although we did not find a significant group difference, our results showed a close correspondence of between the model simulations and the real EEG data and novel sequences generated based on individual models were highly similar to the original stimulus sequence. This provides a proof of concept for future research and suggests that the HBSL model accurately represented the statistical sequence structure in a similar way as did human listeners.

Diffuse large B-cell lymphoma (DLBCL) is an aggressive form of non-Hodgkin lymphoma with a high recurrence rate. The molecular profiling of DLBCL tumors culminated in several immunohistochemistry algorithms for prognostic stratification. Among those, the Hans classifier is widely used for classifying DLBCL into germinal center B-cell-like (GCB) and non-germinal center/activated B-cell-like (non-GCB/ABC) subtypes. The Hans classifier primarily evaluates protein expression of tumor-associated markers, however the tumor microenvironment (TME) of DLBCL includes a myriad of immune and stromal cells, cytokines, and extracellular matrix components that contribute to tumor growth, immune evasion, and recurrence rate. Although the Hans classifier provides a practical method for subtype identification, incorporation of TME information may improve risk stratification and further refine patient groups. Here, we present an unbiased deep learning-based approach to extract meaningful features from TME of DLBCL tumors for the automated processing and analysis of multiplexed images of a DLBCL patient cohort. Our pipeline quantifies a range of features that describe tumor sample cell composition, morphology, and its spatial organization. We point to alterations in the proportions of several cell populations between GCB and ABC tumors including increased immune cell proportions of the ABC and its preferential interaction with the M2-macrophages. Our analysis offers an in-depth characterization of the DLBCL subtypes and is exemplary of how our pipeline can be used for detailed quantitative analysis of a tumor and its subtypes.

We extend the standard susceptible-infected-recovered framework to incorporate natural immune boosting during a short-scale outbreak. By deriving closed-form final size relations, we analytically link total attack rates to boosting dynamics and vaccine coverage. This framework identifies a critical boosting threshold: above it, higher vaccine coverage paradoxically decreases relative vaccine effectiveness. This occurs because successful epidemic suppression deprives vaccinated individuals of the silent pathogen exposures required to maintain their relative immunological advantage. Crucially, the overall population-level impact remains beneficial, consistently reducing absolute disease burden. For highly transmissible variants, asymptotic analysis reveals that relative vaccine effectiveness converges to a positive limit entirely independent of coverage.

Information Processing Pathway Maps (IPPMs) offer a scalable framework for formalizing the complex sequence of mathematical transformations applied to sensory stimuli. These maps chart the latency and cortical expression of computational steps, relying on statistical inference to link model outputs with observed neural activity. Traditionally, this mapping has relied on frequentist hypothesis testing. However, determining which of several competing computational models best explains neural data is a problem of model adjudication, arguably better suited to probabilistic inference. Here, we present a direct comparison between the established frequentist approach and a novel Bayesian framework for mapping cortical entrainment. While the Bayesian formulation retains the core strength of IPPMs -- generating explicit predictions of time-varying neural signals -- it fundamentally alters the selection criterion, shifting from rejecting a null hypothesis to quantifying the relative evidence for competing computational hypotheses. We evaluate the performance and interpretability of both approaches using an auditory neuroimaging dataset to reconstruct a known loudness-processing pathway. We discuss the implications of this shift for systems neuroscience, specifically regarding the handling of collinear models and the robust accumulation of evidence.

Many all-atom generative protein models can in principle design large multimeric complexes by jointly modelling all chains, but their quadratic token- and atom-pair representations quickly exceed single-GPU memory as the number of chains and residues modelled grows. We introduce Design-CP, two context-parallel (CP) inference strategies for RFdiffusion 3 (1D row-sharding and 2D grid sharding with ring attention) that distribute the quadratic activations across a multi-GPU mesh while preserving pretrained weights. We characterise their scaling when sampling icosahedral assemblies, showing that the maximum feasible asymmetric subunit (ASU) size grows with the expected square-root trend in GPU count and that 2D sharding achieves better wall-clock scaling. Moreover, we show how strong point-group symmetry constraints make CP usable out of the box for end-to-end, all-atom design of icosahedral nanoparticles, yielding favourable in silico structural and interface metrics. Finally, we demonstrate octahedral nanoparticle design on a small cluster of workstation-grade 16GB GPUs, illustrating how Design-CP can be a practical path towards democratising large-assembly protein design.

While recent advances in large language models have enabled end-to-end automated manuscript generation, existing systems suffer from three critical deficiencies: (i) generated claims are not deterministically grounded in verifiable literature, (ii) experimental results are frequently fabricated rather than executed, and (iii) there exists no standardized, multi-dimensional framework to assess whether AI-generated manuscripts meet the quality and rigor required for real-world publication. We present Prompt-to-Paper, a multi-agent framework that directly addresses this evaluation gap through three integrated innovations. First, a deterministic retrieval-augmented generation pipeline with section-aware relevance scoring and snowball citation expansion grounds every claim in a verifiable corpus of 60--100 papers. Second, an autonomous coding agent executes real computational biology experiments replacing synthetic outputs with genuine numerical results. Third, an eight-dimensional automated quality scorer, benchmarked with approximate reference statistics from published papers and augmented with explicit hallucination penalties, provides standardized, reproducible quality assessments. The quality-driven improvement loop uses a context-rich reviser that routes each iteration to one of three researcher actions and fires a deep research cycle every ten iterations to re-run experiments and re-manuscript from stronger outputs. We validate the system on five bioinformatics case studies; all five cases compiled submission-formatted PDFs with zero out-of-range citations. The improvement loop raises manuscript quality by an average of +17.96 points on a 0--100 scale (maximum +26.04. As partial external checks, a human reviewer scored the five manuscripts at an average of 7.0 out of 10. Complete manuscripts are produced at approximately 0.31 USD per paper.

Rules, regulations, and regulatory systems are central to societies, institutions, and organisms, yet surprisingly little is known about their evolution over long timescales. The Laws of Cricket, the world's second most popular sport, offer a unique insight into this fundamental question. Their 268-year history constitutes the longest continuous rule-set record yet assembled. Our quantitative analysis reveals generic features including rule-book size growing exponentially in time but scaling sublinearly with matches played; new situations stimulate new rules, but at a decelerating rate; regulatory structures exhibit abrupt phase transitions, increasing rule specificity, interconnectivity and complexity with central rules shifting from gameplay to officiating. These provide a framework for understanding how governance evolves from simple collections of rules to complex regulatory architectures across social, legal, and biological domains.

Every chemical language model reading SMILES begins with a tokenizer, yet the field has inherited byte-pair encoding (BPE) from natural language with little scrutiny. In natural language, BPE's principal alternative, Unigram-LM, is known to build structurally different vocabularies. Whether that contrast survives in chemistry was open. We report a controlled comparison of BPE and Unigram-LM over a fixed 165-token chemistry base, at the small vocabulary sizes where token embeddings are learnable, across three corpus typologies (diverse, drug-like, natural-products) and both pre-tokenization boundary policies. The two do not converge. In all 22 matched conditions they build near-disjoint subword vocabularies: cross-algorithm Jaccard overlap on the learned pieces never exceeds 0.161, and at most 0.05 once weighted toward the high-frequency pieces a model updates most. Unigram-LM also segments held-out molecules into 29-41% more tokens; the arms largely agree on where to cut but not how deeply, so BPE's segmentation is a strict coarsening of Unigram-LM's on 80-99% of molecules. The separation holds across corpus, boundary, and vocabulary size, persisting even at eight times that scale. The subword algorithm is therefore a modeling decision, not a free default. The study trains no language models.

Large-scale circular gap closure occurs over a time scale on which cell growth and proliferation become important. Growth is the main driver of the closing process, while cell dynamics such as elongation and intercalation reflect elastic and fluidic contributions to tissue deformation. We develop a novel fluidized growth-elasticity framework as a nonlinear analogue of a Maxwell fluid with growth. The framework decomposes the experimentally observable strain rate into the additive sum of the growth, elastic, and fluidic strain rates, thus enabling the separate quantification of these contributions from tissue kinematics and allowing the roles of tissue elasticity and fluidity (the inverse of viscosity) to be characterized. We apply the model to large circular gaps ($\sim$1.7 mm in diameter) in confluent monolayers of mouse embryonic epicardial cells (MEC1) under two conditions, without and with TGF-$\beta$ treatment. We show that both tissue fluidity and the elastic properties associated with fiber reinforcement are critical for reproducing the closure kinematics. Specifically, we predict that the treated condition has lower fluidity, associated with a lower fluidic deformation rate and a higher elastic deformation rate than the untreated condition, in agreement with the experimental observations.

Accurate ranking of antibody candidates according to their binding affinity is essential for therapeutic antibody discovery. However, existing methods treat affinity comparisons independently and ignore the contextual information encoded in other labeled comparisons, limiting their ability to capture antigen-specific binding landscapes. For many target antigens, a small number of experimentally characterized affinity comparisons are often available. An important question is whether the model can exploit these existing comparisons to infer antigen-specific ranking patterns that facilitate subsequent affinity ranking. This form of learning from labeled demonstrations closely resembles the paradigm of In-Context Learning, motivating us to revisit antibody affinity ranking from an ICL perspective. To this end, we propose AbICL, an ICL framework for antigen-specific antibody affinity ranking. AbICL combines a pretrained structural encoder with a context ranking head and is trained with an episodic meta-training strategy that enables the model to leverage support demonstrations for test-time adaptation without gradient updates. Experiments on the AbRank benchmark demonstrate that AbICL consistently outperforms existing ranking baselines across almost all data splits and evaluation benchmarks. Further analysis shows that the value of contextual demonstrations depends on how well they match the target inference task, and becomes increasingly pronounced under distribution shift and fine-grained affinity discrimination. These findings highlight the potential of ICL as an effective paradigm for antigen-specific antibody affinity ranking, particularly in challenging settings where a single global ranking function is insufficient.

Computational modeling and simulation are powerful tools for the assessment of medical device performance and safety, particularly for in silico clinical trials for automated medical systems. In ventilation, where managing gas exchange, respiratory mechanics, and patient-ventilator interaction is required under evolving pathophysiology, the clinical translation of automated control strategies remains slow and resource-intensive. This paper applies a standards-aligned framework for the credibility assessment of a computational respiratory model, demonstrated using an automated weaning case study. The framework operationalizes ASME V&V 40 and FDA principles within a structured, guidance-based validation workflow. The computational physiological model integrates respiratory mechanics, gas exchange, respiratory control, and a ventilator representation, validated under a clearly defined context of use and explicit questions of interest. Model credibility is assessed through calibration, physiological plausibility, population-based evaluation, and reproduction of emergent behavior. All model requirements derived from the intended context of use are addressed within the proposed credibility assessment plan, and documented gaps are transparently reported. The resulting credibility argument supports the applicability of the model for its context of use. Strengths are demonstrated in population-based comparison and mechanistic plausibility, while residual limitations relate to the extent of in vivo evidence, population representativeness, and external validation. Overall, the model is considered fit for purpose for medium-low risk preclinical in silico clinical trials of automated weaning strategies. Furthermore, the validation procedure outlined in this article provides a blueprint for the validation of this and similar models in other mechanical ventilation algorithms and related use cases.

Many bioinformatics algorithms, such as sequence alignment and structure prediction, can be expressed as recurrence equations over a dynamic programming matrix. Efficient implementations of these algorithms for large-scale biological data often require changing the order in which matrix cells are calculated and pruning ineffectual regions of the matrix from consideration altogether, but these techniques typically complicate implementation. We introduce FILTR, a domain-specific language (DSL) and compiler framework for bioinformatics recurrences. FILTR keeps the core recurrence rules separate from the pruning and scheduling strategies, where pruning acts as an approximation to limit where in the DP matrix cells are computed, and scheduling determines the iteration order for how cells are explored. FILTR compiles these high-level descriptions into optimized C++ code that matches the performance of hand-tuned implementations while enabling rapid exploration of new heuristics. FILTR is competitive with hand-optimized sequence-alignment libraries, ranging from 0.95x to 30x faster across biological benchmarks.

We introduce EntroPath, a manifold learning method that recovers geodesic geometry from data graphs through ensembles of diffusion paths. Many existing graph-based embeddings rely either on locally normalised random walks or on shortest-path distances. The former can concentrate diffusion in densely sampled regions, while the latter are sensitive to spurious shortcut edges in the graph. EntroPath instead builds its dissimilarities from the maximum entropy random walk (MERW), which aggregates the full ensemble of k-step paths between points rather than relying on any single trajectory. We show that the resulting free-energy dissimilarity converges to squared geodesic distance in the short-time limit, via Varadhan's heat-kernel formula. The diffusion depth k interpolates smoothly between local neighbourhood structure and global manifold geometry, and the symmetrised kernel admits an exact Gram factorisation connecting EntroPath to kernel methods. We further provide scalable extensions via landmark projection and diffusion-potential pseudotime. Across synthetic manifolds and single-cell benchmarks, EntroPath consistently matches or outperforms diffusion- and shortest-path-based methods, while remaining competitive with neighbourhood-preserving embeddings (UMAP, t-SNE) on local-structure metrics. Its gains are most pronounced on manifolds with non-uniform sampling density and well-separated branching trajectories, where path-ensemble diffusion more faithfully preserves the underlying geodesic geometry.

Aging includes both continuous gradual decline, such as in physiological function, together with major deficit onset events such as morbidity, disability and ultimately death. These deficit events are stochastic and include non-linear feedbacks, making health trajectory forecasting challenging. We propose a framework for modelling the gradual effects of aging together with health deficit onset events, as reflected in the frailty index (FI) - a quantitative measure of overall age-related health. We model damage and repair dynamics of the FI from individual health transitions within two large longitudinal studies of aging health, the Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA), which together included N = 47592 individuals. We find that both damage resistance (robustness) and damage recovery (resilience) rates decline smoothly with both increasing age and with increasing FI, for both sexes. This leads to two distinct dynamical states: a robust and resilient young state of stable good health (low FI) and an older state that drifts towards poor health (high FI). These two health states are separated by a sharp transition near age 75. Since FI accumulation risk accelerates dramatically across this tipping point, ages 70-80 are crucial for understanding and forecasting late-life decline in health.

Tree balance has received considerable attention in recent years, both in phylogenetics and in other areas. Numerous (im)balance indices have been proposed to quantify the (im)balance of rooted trees. A recent comprehensive survey summarized this literature and showed that many existing indices are based on similar underlying principles. To unify these approaches, three general metaconcepts were introduced, providing a framework to classify, analyze, and extend imbalance indices. In this context, a metaconcept is a function $\Phi_f$ that depends on another function $f$ capturing some aspect of tree shape. In this manuscript, we extend this line of research by introducing a new metaconcept based on the heights of the pending subtrees of all inner vertices. We provide a thorough analysis of this metaconcept and use it to answer open questions concerning the well-known $B_1$ balance index. In particular, we characterize the tree shapes that maximize the $B_1$ index in two cases: (i) arbitrary rooted trees and (ii) binary rooted trees. For both cases, we also determine the corresponding maximum values of the index. Finally, while the $B_1$ index is induced by a so-called third-order metaconcept, we explicitly introduce three new (im)balance indices derived from the first- and second-order height metaconcepts, respectively, thereby demonstrating that pending subtree heights give rise to a variety of novel (im)balance indices.

Clinical interpretation often assumes that observable performance sufficiently reflects the organization of an adaptive system. However, similar observable performance may arise from distinct latent organizations. This study extends a previous multi-level framework by introducing Level 4, centered on observed longitudinal viability. Using an exploratory single-case design, gait data from a Parkinsonian participant were recorded with instrumented insoles under three occlusal conditions: neutral natural occlusion (ONL), a 2.5-degree increase in vertical dimension of occlusion (OC2.5), and a 3-degree increase (OC3). Two sessions, separated by eleven weeks and a structured sensorimotor intervention, were analyzed. The vertical dimension of occlusion was treated as an experimentally varied constraint applied to an adaptive neuromechanical system. Although observable performance remained broadly comparable across conditions, PCA-based latent-space analysis revealed distinct longitudinal centroid displacements: OC3 showed the smallest displacement, ONL an intermediate displacement, and OC2.5 the largest. Bootstrap resampling confirmed the stability of this Euclidean ordering within the dataset. A complementary Mahalanobis analysis, accounting for within-condition covariance, produced a different ranking, indicating that covariance contributes to the observed displacement. Rather than invalidating Level 4, this finding shows that the proposed viability measure should be interpreted as an exploratory observational proxy rather than as a covariance-independent metric or validated biomarker. These within-subject, exploratory, and non-causal results suggest that clinical relevance depends not only on instantaneous performance or latent organization, but also on the capacity of a configuration to maintain coherent longitudinal behavior over time.

Reticulate evolutionary events, such as hybridization, recombination, and horizontal transfer, can make a tree model inadequate. When evolutionary data are summarized as hardwired clusters, one can ask how much local reticulation complexity is forced by the data itself. We address this question for an arbitrary cluster system $\mathcal C$ on a finite taxon set $X$ by computing the minimum level of a rooted phylogenetic network whose hardwired cluster system is exactly $\mathcal C$. Writing $H=\mathcal H[\mathcal C]$, we define for each non-trivial block $B$ of $H$ a parameter $\mu(B)$ from generating sets of incompatibility intersections in $B$. If $\ell(\mathcal C)$ denotes the minimum level of any rooted network $N$ with $C_N=\mathcal C$, then \[ \ell(\mathcal C)=\max\{\,\mu(B)\mid B\text{ is a non-trivial block of }H\,\}. \] Equivalently, $\mathcal C$ is realizable by a rooted level-$k$ network if and only if $\mu(B)\le k$ for every non-trivial block $B$ of $H$. The lower-bound proof relates incompatibility intersections to non-root hybrid vertices in realizing blocks, while the upper-bound proof starts from the Hasse diagram and iteratively splits selected hybrid vertices without changing the hardwired cluster system. The result turns a network-design problem into a cluster-side criterion and provides an interpretable complexity score for hardwired cluster data, distinct from softwired cluster representation where clusters need only occur in one displayed tree.

Nitrous oxide (N2O) emissions from water resource recovery facilities (WRRFs) fluctuate over time and can arise from multiple microbial pathways, making source attribution and full-scale prediction difficult. The difficulty is compounded by the high dimensionality of activated sludge microbiomes, whose complex and dynamic community structure can obscure relationships with N2O emission patterns. This study evaluated whether interpretable, low-dimensional representations of activated sludge microbiomes can be correlated with N2O emission states. Temporal 16S rRNA gene amplicon profiles and N2O emission metrics were collected from two full-scale WRRFs in Switzerland. Genus-level relative-abundance profiles were summarized using archetypal analysis (AA), which represents each sample as a convex combination of a small number of interpretable community profiles. In both WRRFs, three archetypes captured most explainable variation in community composition (63%--73%) and defined a simplex state space in which samples clustered near vertices and edges, indicating that community compositions were organized around distinct archetypal states and their mixtures. Without using emission labels while training, the archetypal state space aligned strongly with binary N2O emission states: high-emission observations in both plants concentrated around a specific archetype, and temporal trajectories showed consistent high weights of this archetype during high-emission periods. Functional summaries suggested site-specific but pathway-relevant interpretations of the high-N2O archetype. Temperature further structured the archetypal state space, indicating seasonal forcing of microbiome configurations associated with elevated N2O. Overall, AA provides an interpretable framework to track microbiome regime shifts and may support operational tracking of high-N2O emission states in full-scale WRRFs.

This study presents a comprehensive analysis of bird diversity across Sri Lanka by integrating spatial, temporal, and environmental data. Bird observation records were combined with environmental variables, including weather conditions, air pollution, the Normalized Difference Vegetation Index (NDVI), land cover, elevation, and Artificial Light At Night (ALAN), and rigorously preprocessed to ensure data quality. Spatial analyses were conducted on multiple grid scales (2 km, 5 km, 10 km) to evaluate patterns in species richness while minimizing sampling bias through spatial thinning. Temporal trends were assessed using effort-corrected metrics including rarefied richness and occupancy rates to account for variations in observation effort over time. Environmental drivers of bird diversity were examined using multivariate statistical models, including Poisson Generalized Linear Models (GLMs) and correlation analyses, to identify key associations between ecological factors and species richness. Additionally, community structure, dominance patterns, and beta diversity were analyzed to understand variations in species composition across regions and time. The study found that land-cover type is a stronger predictor of bird diversity than individual continuous variables such as NDVI or temperature alone. Urbanization, measured by ALAN, exhibits nuanced scale-dependent effects, supporting high abundances of a few generalist species while reducing overall richness. The findings provide actionable insights into the patterns and drivers of avian diversity in Sri Lanka, offering a scalable and reproducible framework for biodiversity research and conservation planning.

Representational similarity analysis and related methods compare the internal geometries of neural networks, but they measure only alignment between spaces, leaving a blind spot -- whether a representation's structure is reliably recoverable, not merely similar. We introduce geometric stability, a distinct axis, and \textit{Shesha}, a metric that quantifies it from a single representation by correlating dissimilarity matrices built from complementary random halves of the feature dimensions. Unlike CKA and Procrustes distance, Shesha is provably non-invariant to orthogonal rotations of the feature basis. This is by design: the basis is privileged for learned models, since probes, patching, and steering act on coordinates, and a rotation-invariant metric cannot see whether the targeted structure survives them. A double dissociation isolates the mechanism -- removing the top principal component collapses CKA while Shesha holds, whereas rotating a representation into its eigenbasis, which preserves the spectrum and CKA exactly, collapses Shesha. Across 2,463 encoder configurations in seven domains, the metrics are redundant under geometry-preserving transforms and anti-correlate under compression ($\rho=-0.47$). Across 170 vision models spanning 6 clean and 38 corruption-shifted datasets, DINOv2 ranks first or second in transferability on three of six clean datasets yet bottom-quartile in stability on five, an isolated dissociation rather than a trade-off.

Dynamic models are widely used to mathematically describe biological phenomena that evolve over time. One important area of application is leukaemia research, where leukaemia cells are genetically modified in preclinical studies to explore new therapeutic targets for reducing leukaemic burden. In advanced experiments, these studies are often conducted in mice and generate time-resolved data, the analysis of which may reveal growth-inhibiting effects of the investigated gene modifications. However, the experimental data is oftentimes evaluated using statistical tests which compare measurements from only two different time points. This approach does not only reduce the time series to two instances but also neglects biological knowledge about cell mechanisms. Such knowledge, translated into mathematical models, expands the power to investigate and understand effects of modifications on underlying mechanisms based on experimental data. We utilise two population growth models -- an exponential and a logistic growth model -- to capture cell dynamics over the whole experimental time horizon and to consider all measurement times jointly. This approach enables us to derive modification effects from estimated model parameters. We demonstrate that the exponential and logistic growth model recognise simulated scenarios more reliably than a statistical test. Moreover, we apply the population growth models to evaluate the efficacy of candidate gene knockouts in patient-derived xenograft models of acute leukaemia.