Ecological momentary assessment (EMA) is widely used to collect repeated self-report data in participants' everyday lives using mobile devices. EMA studies often involve multiple questionnaires, flexible schedules, and longitudinal data collection, requiring reliable systems for study setup, deployment, monitoring, and data management. Existing workflows are often fragmented across tools, making studies difficult to reproduce and maintain. We present momenTUM, an open-source platform for designing, deploying, and managing mobile EMA studies. Its central principle is that a structured study specification serves as the shared representation across the full workflow. The same specification supports authoring in the Study Designer, execution in the participant-facing mobile application, backend synchronization and storage, REDCap-linked data handling, and researcher monitoring. This makes protocols reusable, inspectable, and consistent across system components without requiring study-specific app implementations. momenTUM integrates with REDCap to automate project setup and synchronize responses. Its authenticated dashboard provides tabular and calendar-based views, filtering by study components, and visual summaries. The platform has been deployed in real-world studies, including AMBIENT-BD, which examines mood, sleep, and circadian rhythms in bipolar disorder, and the EcoSleep cohort study. We also describe an exploratory LLM-assisted extension that generates draft structured study specifications for researcher review. These deployments show that momenTUM can support complex ambulatory assessment protocols while reducing technical overhead and enabling reproducible, reusable, and extensible EMA workflows.

Humans recognize movements effortlessly, even from noisy and complex visual input. But what information in the stimulus allows humans to rapidly classify movements? No framework has systematically compared different strategies of movement analysis to address this question. Here, we used videos of 16 daily activities from the MoVi dataset and compared three strategies: Temporal Movement Primitives (TMPs), which decompose movements into weighted sums of temporally smooth basis functions; Legendre polynomial coefficients, which project joint-coordinate trajectories onto an orthogonal polynomial basis; and Autoencoder latent embeddings. Legendre coefficients and TMPs achieved the highest classifier accuracy, followed by autoencoders. We found two discriminative features for movement classification. The most informative is the general posture of the body, the average spatial configuration that distinguishes one activity from another. Additionally, we identified 9 critical joints that are most predictive for movement classification. Interestingly, good classification accuracy did not automatically lead to good movement generation: when we reconstructed movements for each activity, TMPs preserved the temporal dynamics and produced perceptually natural motion, whereas reconstructions from Legendre coefficients retained only the average posture and appeared frozen. These results reveal a dissociation in how movement information is organized: the static configuration of the body suffices to classify what activity is performed, but the temporal dynamics of movement are required to reconstruct how it unfolds. This distinction clarifies which features the visual system may rely upon for rapid action recognition, and suggests that postural features could enable efficient movement screening in clinical applications, while dynamic information remain essential wherever movement generation is the goal.

Braga and Wardil [J. Phys. A: Math. Theor. 55 (2022) 025601] introduced a population model for two prey species that compete among themselves and are preyed upon by a single predator species. They showed the existence of 16 dynamic scenarios and stated sufficient conditions for the stable coexistence of the three species. The model, which can be seen as replicator dynamics with predator-dependent fitnesses for the prey, is based on two pay-off matrices: one for prey reproduction and one for interaction between prey and predators. We argue that the model can also be seen as a Lotka-Volterra-type predator-prey model with a single prey species, logistic limitation for prey, and frequency-dependent reproduction and capture coefficients. Using this alternative viewpoint, we obtain conditions for the existence of equilibria with the three types of individuals. We also prove theorems on the stability or instability of equilibria with only two species and relate the stability change of these equilibria to the appearance or disappearance of equilibria with the three species. When all parameters, except the one that regulates carrying capacities, are fixed, a rich cascade of bifurcations may appear. Solutions range from predator extinction due to insufficient prey to predators coexisting with one or two prey types. Sometimes stable limit cycles involving all species appear.

Recent advances in generative and embodied AI have been driven by large-scale predictive learning over multimodal data. However, the resulting systems remain largely based on passive training regimes where linguistic regularities create the scaffold onto which information from other modalities is attached. Conversely, neuroscience and cognitive science suggest that biological intelligence is organized in the opposite way, where grounded world models acquired through interaction with the environment provide the semantic scaffold to which language is attached. Here, we illustrate five examples of neural circuits supporting grounded world modelling, which underlie navigation in physical and conceptual spaces, affordance-based perception and interaction with objects, active perception and exploratory learning, allostatic control and emotion, and the distinction between self- and world-generated outcomes. These examples highlight several features largely missing from current embodied AI, including the role of intrinsic dynamics as a foundation for learning, the centrality of action in aligning these dynamics with the external world, the prominence of autonomous experience and open-ended learning over passive assimilation of externally provided data, and the fact that early predictive and control mechanisms scaffold higher cognitive abilities such as reasoning, conceptual navigation, planning, imagination, understanding others' minds, and communication. Finally, we discuss whether and how principles derived from biological systems may inform future embodied AI, including training regimes based on social interaction to construct world models that are not only grounded but also socially shared and aligned with human norms and values.

Next-generation sequencing technologies, including RNA-sequencing, provide genome-wide measurements of gene expression and enable broad explorations of biomarkers and mechanisms underlying disease and treatment response. Bioinformatics tools for processing this data, such as differential expression analysis, are largely univariate, linear, and rely on predefined pathway knowledge annotations, which limits their ability to capture nonlinear and multivariate gene interactions. This paper explores the application of causal discovery to characterizing transcriptional responses to radiation as a function of dose rate in human cells. By jointly modeling radiation perturbations and gene expression, we learn directed gene networks that capture important regulatory relationships beyond correlation and exhibit significant enrichment of known radiation response pathways compared to baseline approaches. We find that inferred causal graphs reveal structured network features such as high in-degree housekeeping genes and high out-degree transcription factors. Further analysis suggests a hierarchical organization of stress response pathways and triggered cell death pathways. This work highlights the potential of causal discovery in healthcare settings with applications to understanding response mechanisms, identifying regulatory targets, and improving interpretation of complex genomic data.

Transient synaptic memory has emerged as a potential mechanism for maintaining short-term information even in the absence of persistent neuronal activity. However, it remains unclear whether the hidden synaptic state alone contains sufficient information to predict the future evolution of neuronal networks after activity has ceased. Here, we introduce a minimal neuronal network model with finite-lifetime synapses and investigate the mechanism underlying spontaneous activity regeneration following complete neuronal silence. We show that the residual synaptic configuration at the first silent state already determines whether network activity terminates after a single activation cycle or spontaneously regenerates an additional cycle. By analyzing this synaptic-memory snapshot, we identify the Latent Excitatory Recruitment (LER) capacity, quantified by the cumulative number of fresh excitatory neurons, as a near-perfect predictor of multi-cycle dynamics without continuing the subsequent network simulation. Remarkably, these distinct dynamical outcomes emerge in an otherwise homogeneous neuronal network, demonstrating that transient synaptic memory alone is sufficient to generate diverse future dynamics. Our findings provide a mechanistic explanation for activity regeneration from a residual synaptic state and suggest that short-term memory is encoded not only in ongoing neuronal activity but also in the latent synaptic configuration that preserves the network's capacity to recruit new neuronal assemblies. More broadly, the proposed snapshot-based framework offers a new perspective for predicting and potentially controlling the future evolution of neuronal networks.

Genomic foundation models such as Evo 2 learn rich sequence representations, but their value for biosecurity screening is largely unexplored. We ask how much biosecurity-relevant signal is linearly accessible in these representations by training minimal linear and attention probes on frozen Evo 2 layer-26 activations, without fine-tuning the underlying model. Across held-out metagenomic test sets, the probes detect antimicrobial resistance (AMR) with strong discrimination: a linear probe reaches a region-level ROC-AUC of 0.888 (mean-pool), rising to 0.977 with a single-head attention probe. The probes resolve finer-grained AMR drug-class subcategories and separate them from unrelated functional genes, providing additional evidence that the learned signal is not explained solely by generic functional-gene status. Bacterial virulence is also decodable, though more weakly (region-level ROC-AUC 0.833). The AMR probe retains comparable ranking performance on simulated short reads without retraining, enabling evaluation before assembly in settings where assembly is computationally costly or unreliable. It achieves a read-level ROC-AUC of 0.898 (mean-pool), comparable to the mean-pooled full-region result. Within SynGenome, AMR-associated prompt labels are only weakly recoverable from Evo 1.5-generated sequences; these prompt-derived labels do not establish the function of the generated response sequences. A complementary sparse-autoencoder analysis recovers interpretable resistance-associated features but proves less consistent than the supervised probes. Together, these results position lightweight embedding-based probes as a fast, inexpensive first-pass detection layer for metagenomic biosurveillance and map both strengths and current limits of the approach. This work was conducted as part of the AIxBio Hackathon 2026 hosted by BlueDot Impact, Apart Research, and Cambridge Biosecurity Hub.

The finite-temperature conductance of a molecule coupled to metallic leads is derived entirely within the framework of density functional theory (DFT) and its time-dependent extension for open quantum systems. Starting from the Mermin grand potential, the foundational Kohn-Sham equations, the Fukui function, and the open-system master equation for the single-particle density matrix are systematically formulated. The non-equilibrium electron-phonon dissipator is obtained from the partial trace over the phonon bath. By applying Wick's theorem for non-interacting fermions, a fully exchange-symmetric collision integral is obtained that strictly preserves Pauli exclusion at the operator level. Performing a double perturbation expansion, initially in the applied voltage (linear response), and subsequently in the molecule-lead coupling (weak coupling), it is demonstrated that under the fast-thermalization condition, the complex exchange-correlation self-consistent field response is analytically projected out by the diagonal structure of the slow Liouvillian mode. Consequently, the thermal conductance is governed by the finite-temperature Fukui function, the central reactivity descriptor of conceptual density functional theory. This condition is satisfied in proteins, whose wave functions are extended and multifractal due to quantum criticality at the Anderson metal-insulator transition. This derivation establishes a fundamental link between electronic transport and chemical reactivity, identifying conducting paths with reactive sites. It opens new technological avenues connecting drug design to conductance experiments and also provides a foundation for designing next-generation bioelectronic sensing and computing architectures.

Interfacing with Biological Neural Networks (BNNs) requires encoding information into stimulation patterns that can be effectively processed and that enable the underlying system to adapt. Nevertheless, the role of stimulation encoding remains poorly understood. In this work, we compare multiple encoding strategies, including rate-based, phase-based, burst-based, and time-to-first-spike temporal encodings, in a closed-loop neural classification task using cultured BNNs. We encode visual inputs as spatiotemporal stimulation patterns delivered via a Multi-Electrode Array (MEA) and evaluate classification performance for each encoding scheme. We find that burst-based temporal encoding yields the highest observed performance, achieving up to 95.6 % accuracy in a binary classification task, compared to substantially lower performance from rate- and phase-based approaches. We further show that performance is highly sensitive to the spatial distribution of stimulation, with suboptimal electrode selection significantly degrading accuracy. These findings indicate that effective interfacing with biological neural systems requires the joint optimization of temporal and spatial encoding strategies, and highlight temporal encoding as a key design dimension for bio-digital computing.

Hamilton-Jacobi Reachability (HJR) is a central framework in safe control theory. While HJR has traditionally focused on a few fundamental tasks, there is increasing interest in scaling to more complex objectives. Recent works have studied the exact decomposition of the value functions for two fundamental dual-objective tasks in the adversary-free setting. However, not all value function decompositions in HJR remain valid with an adversary. In this work, we develop theoretical approaches to certify that for these two composite value functions, the proposed decompositions still hold with an adversary. Finally, we show how these results can solve issues that arise when applying HJR to optimal drug regimen design.

Human-centered adaptive systems require behavioral models that are both psychologically interpretable and mathematically analyzable. Many existing predictors either operate as black-box input-output mappings or provide limited access to latent internal dynamics. This paper addresses this gap by modeling behavior as a perception-cognition-decision pipeline. We propose a modular state-space model in which attentional selection, predictive inference, cognitive-state evolution, intention formation, and action selection are represented by coupled mathematical mappings. The model links sensory inputs to observable behavior through latent internal states while retaining interpretable connections to neuro-cognitive mechanisms. We establish sufficient conditions for boundedness, Lipschitz regularity, forward invariance, contraction of perceptual inference under constant input, and input-to-state stability of the cognitive state dynamics. Numerical sensitivity analyses show that the model yields interpretable changes in perceptual tracking, cognitive amplification, intention expression, and action decisiveness. We further demonstrate a closed-loop rehabilitation case study in which a receding-horizon controller uses the model to adapt movement difficulty from partial feedback. In this proof-of-concept setting, the model-based controller sustains simulated task participation and achieves lower realized cumulative cost than target-following and random baselines. Overall, the framework provides a white-box dynamical structure for estimation, validation, and model-based control in human-centered settings.

Robust and accurate neural decoders are integral to neurotechnologies such as brain-computer interfaces and closed-loop experiments. Recent work has shown that tokenizing neural data at the spike level facilitates multi-session pretraining and delivers state-of-the-art decoding performance. However, current spike-based models are restricted to supervised learning (SL), limiting training to datasets with paired behavioural labels. To address this limitation, we introduce MOJO (Masked autOencoder-based JOint training), a training framework for spike-tokenizing models that jointly leverages self-supervised learning (SSL) via masked autoencoding and SL objectives. We evaluate MOJO on three spiking datasets spanning monkey motor cortex during reaching tasks and multi-regional mouse recordings during vision and decision making tasks, demonstrating superior performance over purely SL-trained models. This improvement is especially pronounced when training with limited labelled data, particularly in few-shot finetuning, where only a small amount of labelled data from a new session is available. Incorporating SSL also yields more interpretable neuronal representations, improving performance on brain region classification and spike-statistics prediction without explicit optimization for these tasks. We further show that MOJO generalizes beyond spiking data to human electrocorticography during speech, where it continues to outperform purely SL-trained models and achieves performance comparable to neuro-foundation models (NFMs) designed specifically for continuous signals. Overall, augmenting spike-tokenizing models with SSL improves performance in label-impoverished settings and enables the use of unlabelled data across various tasks and species, while generalizing to other neural modalities. These results suggest a path towards more flexible and scalable data usage when training NFMs.

Evolution is the adaptation of populations to their environment expressed through the concept of fitness. Darwin did not define fitness but described evolution as the higher prevalence of lineages with advantages in survival and reproduction in changing environments, an implicitly statistical and relational notion. As evolutionary dynamics became more quantitative, however, fitness acquired a narrower meaning of relative reproductive success. Crucially, this narrower definition suffers from three fundamental difficulties, known as the circularity, mismatch, and prediction problems. We show that interpreting evolutionary dynamics in terms of inference resolves these three problems while also creating new productive analytical tools. This shift redefines fitness via a Bayesian likelihood, a predictive probability of the environment specific to each type. We show that averaging the growth rate over environmental histories connects selection to information as types with better environmental models are amplified. It follows that long-run evolutionary dynamics maximizes the mutual information between population structure and environmental statistics, establishing information maximization as the governing principle of natural selection. We illustrate this approach in several population dynamics problems including task switching, evolutionary games, and selection in group-structured populations. In each case, we derive phase diagrams as functions of environmental statistics and Hamilton-type rules for the emergence of cooperation, while also demonstrating the generality of the approach.

In response to increasing threats to biodiversity, conservation objectives have been set to halt biodiversity decline by reducing direct anthropogenic drivers. However, the potential effects of these objectives on common species remain rarely studied. We analyse the effect of a range of drivers related to climate, land use and land use intensity, on 265 common bird and 144 common butterfly species from more than 20,000 sites between 2000 and 2021 across 26 European countries. We use land-use and land-use intensity scenarios produced previously using the IPBES Nature Futures Framework, and climate change scenarios in order to project biodiversity drivers in Europe up to 2050. We translate these driver changes into abundance variations for common bird and butterfly species, and for multi-species indicators used to monitor common biodiversity status in Europe. The projected trends relatively improve, while still declining for birds, notably farmland species, under the scenarios meeting conservation objectives, with few effects on butterflies. No scenario shows a stop or a reversal in the average decline in abundance of bird and butterfly species. Our results therefore question the common biodiversity future under current conservation policies and highlight the need for other anticipatory frameworks, not implicitly based on a growing need for natural resources.

Detecting alternative splicing from complex splicing bubbles that contain more than two transcript paths are challenging. We present GrASE, a framework that distinguishes alternative path groups by the distinct and shared exonic parts they contain. Using lower-set bipartitioning, which efficiently enumerates downward-closed sets of paths, we find bipartitions of complex bubbles with well-defined distinct exonic parts. Applied to the human genome, GrASE tests 496,565 path groups from valid bipartitions across 22,299 genes, including alternative TSS and TTS events that are inaccessible to junction-based methods. GrASE can trade recall for higher precision and still detect substantially more events, due to its expanded test universe.

Mass-action networks are special cases of chemical reaction networks. For these systems, we argue that conserved quantities are dual to internal cycles. We introduce maximal invariant polyhedral supports, and we conjecture that there is a duality relation between preclusters and maximal invariant polyhedral supports. Given the close relation between maximal invariant polyhedral supports and siphons, we also conjecture that siphons and preclusters are dual objects.

We mathematically model the dynamics of the number of migratory fish observed at a fixed location along a river in a random environment. Particularly, as a new approach, we construct a stochastic differential equation that incorporates the influence of environmental factors on the fluctuations in the start and end of migration. The model is a diffusion bridge with a non-Lipschitz diffusion coefficient, called the Cox-Ingersoll-Ross bridge, and has random initial and terminal times arising from time-change, so that the influences of environmental factors can be efficiently incorporated. The well-posedness of the model is first established, which is considered novel and significant in applied mathematics. Second, we estimate the parameters of the model based on the latest multiyear daily data set for the upstream migration of Plecoglossus altivelis altivelis (Ayu) by relying on the hypothesis that water temperature affects the migration of the fish, which has been suggested in existing studies. We also explore the application of the proposed model to the challenging task of analyzing environmental DNA data. This study advances the development of a theory of fish migration that is simple yet can take environmental factors into account.

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.

The push toward large language models for biology (BioLM) has created a need for training corpora that can endow models with a genuine understanding of biology. However, existing biological resources, such as molecular databases, protein repositories, genomic annotations, single-cell atlases, and pathway databases, are scattered across heterogeneous formats and remain unorganized into a cohesive corpus for language model training. We present TheBioCollection, a 52.6B-token pre-training-scale corpus that converts these disparate resources into a unified, training-ready form spanning small molecules, proteins, genomic sequences, cells, and pathways. Beyond consolidating existing data, TheBioCollection enriches each record with tool-computed biological properties and introduces new instruction tasks for capabilities that current corpora barely cover. We pair the corpus with TheBioCollection-Eval, a matched suite probing recognition, generation, and prediction across molecular, protein, genomic, cellular, and cross-domain settings. Holding the base Gravity-16B-A3B architecture fixed, training on TheBioCollection more than doubles its overall score on TheBioCollection-Eval with gains in every domain, while leaving general linguistic ability nearly intact.

When humans and large language models (LLMs) process the same text, activations in the LLMs correlate with brain activity measured, e.g., with functional magnetic resonance imaging (fMRI). Moreover, it has been shown that, as the training of an LLM progresses, the performance in predicting brain activity from its internal activations improves more in the left hemisphere than in the right one. The aim of the present work is to understand which kind of competence acquired by the LLMs underlies the emergence of this left-right asymmetry. Using the OLMo-2 7B language model at various training checkpoints and fMRI data from English participants, we compare the evolution of the left-right asymmetry in the correlation between brain activity and model predictions alongside performance on several benchmarks. We observe that the asymmetry co-emerges with the formal linguistic abilities of the LLM. These abilities are demonstrated in two ways: by the model's capacity to assign a higher probability to an acceptable sentence than to a grammatically unacceptable one within a minimal contrasting pair, and by its ability to produce well-formed text. By contrast, the left-right asymmetry does not align with the performance on arithmetic or Dyck language tasks; nor with text-based tasks involving world knowledge and reasoning. We generalize these results to another family of LLMs (Pythia) and two other languages, French and Chinese. Our observations indicate that the left-right asymmetry in brain predictivity matches the progress in formal linguistic competence.