Papers

Physical Activity (PA) is consistently associated with lower risk of dementia, yet the mechanism linking PA to dementia prevention remain incomopletely understood. Here, we integrate large language model (LLM)-guided causal discovery with mediation analysis in 42,293 older adults aged 60 years or older from the UK Biobank to systematically identify pathways connecting objectively measured moderate-to-vigorous physical activity (MVPA) to dementia risk. Across behavioral, psychological, functional, and clinical domains, causal discovery consistently identified interconnected pathways linking higher MVPA to lower dementia risk through depression, functional capacity, smoking behavior, hypertension, cardiovascular disease, chronic kidney disease, and brain injury. Chain mediation analyses further identified depression as a central pathway, accounting for 15.1% of the overall association between MVPA and dementia risk. Sex-stratified analyses revealed distinct mechanistic patterns, with females exhibiting predominantly metabolic and functional pathways, while males showed behavioral and cardiometabolic cascades involving smoking and cardiovascular disease. Together, these findings suggest that the protective association between PA and dementia is mediated through interconnected behavioral, psychological, and cardiometabolic processes rather than a single pathway. Depression emerged as a prominent and potentially modifiable pathway, highlighting opportunities for integrating dementia prevention strategies that combine PA promotion with mental health and cardiovascular risk management.

Microstate analysis starts with clustering and the meta-criterion is a heuristic to find an optimum cluster number. We address the following questions: (i) what number is found for non-cluster-forming dynamical systems, (ii) do centroids fall in random attractor regions, (iii) do EEG topographies form clusters or a single connected structure in sensor space? We find that (i) the meta-criterion suggests spurious optimal cluster numbers (4-8) with high confidence on different attractor geometries, (ii) cluster centroids reliably fall in the same attractor regions, (iii) topological analysis of dynamical systems and resting-state EEG suggests that all form a single connected structure in their respective phase space, not clusters. We conclude that the meta-criterion should be used with caution and its results should not be taken as representing a ground truth. Cluster numbers deviating from the meta-criterion should not be discarded. A more far-reaching implication is that evidence for the existence of clusters in resting-state EEG data is still lacking. Microstate clustering may correspond to the partitioning of a single connected structure.

The BabyLM challenge measures how much language a model can learn from developmentally-plausible, child-scale data rather than internet-scale corpora, yet prior language models forgo the biological constraints of the neural circuitry that acquires human language: spiking neurons separated into excitatory and inhibitory populations wired by a recurrent lateral connectome. This paper presents URCHIN (Unified Recurrent Connectome with Horizontal Integrate-and-fire Neurons), which applies the Parallelized Hierarchical Connectome Spiking State-space Model (PHCSSM) to language modeling: leaky integrate-and-fire neurons coupled by a Dale's-law lateral connectome resolve each token through a multi-transmission loop that recirculates activity to a fixed point. The instantiation is deliberately minimal: a single horizontal layer of 128 neurons, no attention, and 4.23M parameters. Two implementations share one set of weights and produce identical benchmark scores, so URCHIN is trained once and deployed either way with no conversion step: a parallel state-space model (SSM) scan that is GPU-efficient for training, or an event-driven recurrent spiking neural network (RSNN) with constant-cost inference for CPU or neuromorphic edge deployment. Across all three BabyLM tracks (Strict-100M, Strict-Small, and Multilingual), URCHIN offers a biologically plausible, efficient, and directly deployable reference point.

How can a neural circuit rhythmically burst when none of its constituent neurons can endogenously do so? We address this question through a bottom-up reconstruction of a 4-cell neural circuit modeled after the swim central pattern generator (CPG) of the sea slug Dendronotus iris. We first map the intrinsic regimes of a swim interneuron (SiN) model neuron and show that slow mutual inhibition can generate anti-phase bursting in a half-center oscillator (HCO) assembled from tonic-spiking or quiescent cells. A slow–fast phase-space deconstruction explains this pairwise rhythm as a release mechanism. We then move on to CPG parametrization, in which cells 1 and 2 are quiescent, while cells 3 and 4 are tonic spikers but their isolated HCO can only exhibit tonic spiking or suppression. Bursting therefore does not arise at either the cellular or HCO level. It appears only after the two modules are assembled into the complete 4-cell network, where cross excitation, cross inhibition, and rectified electrical coupling act together to support a robust network rhythm-generation. Event-based symbolic encoding and GPU-parallel parameter sweeps show that this higher-order collective state occupies extended parameter domains rather than a single tuned point, and they quantify how the network interactions reshape those domains and the spike counts per burst. Symbolic sweeps identify the healthy range of activity regimes (periodic sequences) and transition boundaries; Lempel–Ziv complexity is used as a descriptor of aperiodic symbolic output rather than as proof of chaos or dynamical instability. Our results establish a hierarchy of rhythm generation—from intrinsic cell dynamics, through conditional HCO bursting, to bursting that emerges only in the fully assembled 4-cell CPG—and provide experimentally accessible predictions for perturbing its chemical and electrical couplings.

Brain-computer interfaces (BCIs) decode neural activity to restore lost function. Typically, training a high-performance neural decoder requires a large labeled dataset to be collected from every new subject. One way to reduce the labeled data cost is self-supervised pretraining, which learns general neural representations from unlabeled recordings that accumulate across subjects. However, for intracranial electroencephalography (iEEG) recordings, self-supervised learning has been challenging due to differences in contact placement and neuroanatomy between subjects. We propose MAPA, an otherwise vanilla masked autoencoder with two spatial encodings, an anatomical region embedding and a relative positional encoding, which together enable it to learn neural representations that transfer to unseen subjects and across various tasks. MAPA sets a new state of the art across all three regimes of the Neuroprobe benchmark without fine-tuning: within-session, cross-session, and cross-subject. In the cross-subject regime, a linear probe on MAPA's features needs only ${\sim}164$ labeled trials to reach the accuracy that takes 3,500 without pretraining. Our results show that self-supervised pretraining can scale across heterogeneous iEEG recordings and reduce the labeled data needed for accurate decoding in new subjects.

The maintenance of continuous variable information in working memory is thought to rely on persistent patterns of cortical activity. In delayed-estimation tasks, neural activity can form localized activity peaks, or “bumps,” whose positions track the remembered variable. Such activity is well described by continuous-attractor neural field models, but most existing models collapse cortical inhibition into a single homogeneous population. Here, we introduce a stochastic neural field model with distinct excitatory, parvalbumin-expressing (PV), and somatostatin-expressing (SST) populations to examine how inhibitory subtype structure shapes persistent activity. Using a Heaviside firing-rate approximation, we derive stationary bump solutions and reduce their linear stability to separate shifting and scaling modes. We show that population thresholds and inhibitory timescales determine both bump stability and the mechanism by which stability is lost, while inhibitory connection strengths and spatial scales substantially reshape the stable parameter region. In particular, broader SST connectivity promotes stable bump states. Finally, we derive an effective diffusion coefficient for noise-driven bump wandering and show that increasing the SST spatial footprint reduces the rate of memory diffusion. Together, these results demonstrate how inhibitory subtype structure can shape both the deterministic stability and stochastic precision of continuous-attractor memories.

The analysis of neuronal avalanches offers insights into brain dynamics utilizing the framework of criticality, but the reproducibility and comparability of studies are limited by the use of fragmented, lab-specific scripts. To address this issue, we introduce pyAvalanches, an open-source Python package providing a standardized, end-to-end pipeline for avalanche analysis from electrophysiological recordings (e.g., electroencephalography-EEG). Starting from the detection of neuronal avalanches the package provides their core statistical characterization, including size and duration distributions. Beyond this, the main aim of pyAvalanches is to characterize the spatiotemporal organization of activity propagation during avalanches. To this end, the core innovation of pyAvalanches is the compuation of Avalanche Transition Matrices (ATMs) to map spatiotemporal propagation patterns. Building on this, the package derives network-based metrics from the ATMs, bridging the study of the topology and organization of the underlying dynamical interactions with network neuroscience adopting the framework of neuronal avalanches. The entire workflow is encapsulated in a modular and scikit-learn compatible architecture. We demonstrate the utility of pyAvalanches through an illustrative group-level analysis on a public resting-state EEG dataset, comparing propagation patterns across different clinical populations. By providing a user-friendly, tested, and extensible tool, pyAvalanches facilitates reproducible research, enables the development of novel avalanche-based biomarkers, and makes complex avalanche analysis accessible to a broader scientific community. The package is fully documented and distributed via the Python Package Index (PyPI).

Motor control has described the surplus of solutions available to the nervous system as redundancy, a term that names duplication: interchangeable elements, robust to loss but incapable of differential adaptation. Biology has had a second term for twenty-five years. Degeneracy names elements that are not interchangeable and are nonetheless isofunctional with respect to a given output, and it supports adaptability, since non-identical elements necessarily diverge in some context. Circuit neuroscience has relabeled its own results accordingly, while motor control has kept the older vocabulary. Neuromechanical models, by placing a spinal circuit in the loop with a musculoskeletal apparatus, bring the two traditions onto the same class of objects. We restate the Edelman and Tononi distinction for sensorimotor systems and derive an operational requirement: not the existence of multiple solutions, but their divergence in contexts they were not selected for. Three influential studies each meet part of that requirement and none meets all. We then argue that degeneracy and redundancy coexist along the sensorimotor hierarchy in a proportion that varies continuously, and that this proportion is measurable: computing degeneracy twice for the same configuration, once with muscle activation as the output and once with the movement produced, isolates what the musculoskeletal apparatus contributes. Five predictions follow, with the single outcome that would refute the proposal. We set out the adaptations the measurement requires in a nonlinear, non-stationary, closed-loop system, and what changes for motor control once solutions are no longer assumed equivalent: the question shifts from which rule selects a command to what the repertoire of the system still allows.

Multimodal large language models predict brain activity, but brain alignment has been a measurement, not a design tool. We propose the Platonic brain bridge hypothesis: omni models, multimodal large language models that process video, audio and text jointly, converge on brain-like representations usable in both directions. From model to brain, brain-likeness of seven omni models is stable across participants, rises with every input channel in three bases, and our encoders lead the Algonauts 2025 out-of-distribution leaderboard. From brain to model, Brain-MoE fixes the expert partition of a frozen base to the seven networks of human cortex, trains experts on network-labelled Brain-AVQA questions, raises held-out accuracy in all 15 model-benchmark pairs by 6.42 percentage points on average and exceeds capacity-matched random experts in 14. Brain-Scope localizes the correspondence to sparse features whose removal weakens brain prediction. Human brain organization is therefore a usable architectural prior for multimodal large language models.

For infinitely many sizes we exhibit a cycle of binary units, each taking the exclusive-or of itself and its two clockwise neighbours, that is a complex in every state, no proper subsystem being more integrated, and whose integrated information grows as two raised to a power quadratic in the number of units.

Parkinson's disease is clinically and biologically heterogeneous, yet its spatiotemporal progression remains poorly characterized. We present a connectome-constrained disease progression model that jointly estimates subject-specific disease time and data-driven subtypes from longitudinal morphometry. Applied to 85 imaging and clinical biomarkers from the Parkinson's Progressive Markers Initiative (PPMI) cohort, the model recovers four morphologically distinct progression subtypes. We validate the model on a hold-out cross-sectional dataset and benchmark it against SuStaIn under a matched training and validation protocol. Only our method recovers subtypes that correspond significantly to clinical motor subtypes and genetic variants of Parkinson's Disease.

Whether human motor and brain lateralization arises from fundamentally distinct neural architectures or emerges from conserved network dynamics remains a central question at the intersection of network science and neurobiology. Conventional measures of cortical activation often fail to resolve how directed information exchange adapts to manual preference during complex motor tasks. This ambiguity leaves it unclear whether left-handed individuals possess atypical neural organization or follow shared dynamical principles. Here, we combine Permutation Transfer Entropy (PTE) with network-based indices to map directed cortical signal flow using electroencephalography (EEG) during high-precision motor execution (handwriting) in both right- and left-handed individuals. We demonstrate that cortical communication is strictly scale-dependent regardless of handedness: ipsilateral dominance occurs primarily in general, network-wide interactions, whereas inter-hemispheric interactions maintain a contralateral profile. Crucially, left-handed individuals exhibit inter-hemispheric information dynamics that are functionally equivalent to right-handers when executing tasks with the same anatomical hand, refuting assumptions of mirrored or atypical lateralization. By demonstrating that human cortical lateralization relies on dynamical principles conserved across manual preferences, these findings suggest that functional asymmetries reflect adaptive optimizations governed by epigenetic or social flexibility, rather than fixed structural divergences.

fMRI foundation models increasingly aggregate heterogeneous data across brain states, cohorts, and acquisition settings, yet pretraining domains are commonly treated as a flat mixture and downstream tasks are adapted independently. We study whether measured learning relations can organize both stages without modifying the backbone. During pretraining, a lightweight Brain-DiT proxy estimates difficulty and directed facilitation across ten fMRI domains, yielding a priority-guided cumulative domain curriculum combined with high-to-low-noise timestep scheduling and joint consolidation. During adaptation, controlled first- and higher-order transfer across fifteen tasks constructs a directed taskonomy, from which budgeted integer programming (BIP) selects directly supervised source tasks and target-specific routes. The joint priority-domain and high-to-low-timestep curriculum reduces v-NMSE, PSD-NMSE, and FC-MSE by 6.5%, 16.3%, and 10.5%, respectively, relative to uniform sampling over both dimensions, and shows strong downstream performance across six in- and out-of-domain tasks. The taskonomy reveals asymmetric, target-dependent transfer, while exploratory sealed-test evaluation shows larger descriptive gains for BIP policies when higher-order route spaces are available than for matched random controls. Together, these findings support organizing fMRI pretraining and adaptation by measured learning relations rather than treating domains and tasks as independent flat sets.

Locust visual systems have long served as an important biological paradigm for studying looming perception and collision avoidance. Numerous computational models have successfully reproduced the selective responses of Lobula Giant Movement Detector (LGMD) neurons to approaching objects, thereby emulating the fundamental functionality of the biological system. However, existing models remain limited in biological plausibility and robustness when operating in complex and dynamic visual environments. To address these limitations, we propose a biologically plausible neural network for locust-inspired looming detection. The proposed framework incorporates a spatially isotropic sampling strategy that mimics the ommatidial organization of the locust compound eye, a population-voting mechanism inspired by population coding in biological neural systems, and leaky integrate-and-fire neuronal dynamics to replace conventional sigmoid-based membrane activation. Systematic experiments on synthetic stimuli, laboratory sequences, and real-world driving scenarios demonstrate that the proposed model improves robustness under challenging visual conditions while preserving computational efficiency and enhancing biological fidelity. These results highlight the potential of biologically grounded neural computation for robust and efficient collision perception.

Source-level analysis of interictal epileptiform discharges (IEDs) is relevant to presurgical evaluation and treatment planning because it helps characterize where epileptiform activity is likely to arise. Beyond detecting whether an IED is present, this setting requires assigning IED-positive activity to clinically meaningful brain-region categories. This setting is challenging because source-region evidence in short electroencephalography (EEG) windows can be subtle, partial, and affected by subject variability, class imbalance, and imperfect multimodal context. We present EEGBind, an EEG-centric multimodal binding framework for five-class source-level IED classification. EEGBind treats EEG as the primary modality and binds synchronized video-context features around an EEG-centric representation. Instead of relying on early or overly strong multimodal fusion, which may perturb the source-sensitive EEG representation, EEGBind uses video context as auxiliary evidence for robust classification. A view-consistent repair stage is further used to improve hidden-set robustness while preserving the learned source-class boundary. On the NeuroMM 2026 Grand Challenge Track 3 NMM-Source-IED benchmark, EEGBind achieves 0.8395 on weighted-F1 and outperforms strong competitors. These results support EEG-centric multimodal binding as a practical strategy for source-level IED classification. The open-source code is available at https://github.com/HKUSTGZ-ML4Health-Lab/NeuroMM2026_IED_Detection.

Research on adaptive systems has traditionally focused on behavior (what organisms do) and mechanism (how their machinery works). This paper focuses on a third level, computation, which considers what adaptive systems must compute to survive and reproduce. It is proposed that adaptation has its own computational structure, comprising a small set of primitive operations common to all adaptive systems, regardless of their physical form. Six primitives, Arouse, Orient, Valence, Position, Boundary, and Attune, were selected using four criteria: necessity for existence, universality across independently evolved lineages, evolutionary conservation, and irreducibility. From this, two main implications follow. First, in biological systems, the primitives provide a substrate-neutral method for cross-species comparisons, reframing elaborative behaviors like attention, memory, and decision-making as combinations of these computations. Second, in artificial systems, the primitives offer a new way to view current challenges in machine intelligence, such as confabulation, prompt injection, distractibility, reward hacking, and catastrophic forgetting, suggesting that these issues may arise from a lack of these computations. Thus, for biological and artificial systems that persist, no specific physical substrate is necessary; rather, these primitives must be implemented if the systems are to be adaptive. What biology offers to inform machine intelligence, then, is not the brain's neural design but the computational functions it evolved to perform. The Computational Primitives Theory (CPT) presented here is a working hypothesis, intended for further refinement through discussion, empirical testing, and application.

Critical systems sit near boundaries between qualitatively distinct behaviors. When inferring models of neural activity, this proximity to criticality is thought to require the precise tuning of parameters. Here, we show that as the number of neurons increases, criticality can emerge naturally without fine-tuning. When computing observable statistics from parameters (the forward problem), some small regions in parameter space map to large regions in statistics space. These special parameters are precisely those near criticality. Thus, when inferring parameters from experimental measurements (the inverse problem), models concentrate near critical points, and this concentration becomes stronger as the system grows. We illustrate this flow toward criticality across many large-scale recordings in the mouse brain. In the Curie-Weiss model of Ising spins, we find that all of the recordings collapse to a first-order phase transition, despite substantial differences in the underlying systems. Together, these results suggest a resolution to the tension between criticality and fine-tuning in models of neural activity.

Brain-age prediction models are commonly evaluated by predictive accuracy, yet accurate predictions alone do not establish that a model relies on reproducible or neurobiologically supported mechanisms. Post-hoc explanation methods can expose these mechanisms, but existing workflows typically stop at diagnosis or require correction targets to be specified before model analysis. We propose XAI-Refine, an automated explanation-knowledge loop for brain-age prediction from resting-state functional connectivity. At each iteration, XAI-Refine consolidates complementary post-hoc analyses across repeated training runs into reliable, structured model explanations. It converts each reliable explanation into a neutral neurobiological question, retrieves and verifies relevant literature, and compiles the verified evidence into an admissible set in the same typed explanation space. The target for refinement is defined as the minimal projection of the current model explanation onto the admissible set induced by applicable verified knowledge. This revised explanation is then translated into a differentiable constraint while preserving the originating model variable, measurement operator, and applicable scope. Candidate updates are promoted only when multi-seed validation confirms target-directed explanatory movement, predictive performance remains within a prespecified guardrail, and non-target explanatory drift remains bounded. Experiments on functional-connectivity-based brain-age prediction evaluate predictive performance, explanation reliability, literature alignment, and target-specific model revision, illustrating a structured route from post-hoc analysis to evidence-guided model refinement.

Hi-M combines fluorescence in situ hybridization (FISH), automated microfluidics, sequential imaging, and computational chromatin tracing to measure the three-dimensional organization of selected genomic regions in single cells. This chapter describes a Hi-M workflow adapted for cryosections of adult Drosophila melanogaster brains, enabling chromatin tracing while preserving tissue architecture and cell identity. The protocol covers Oligopaint library design and amplification, fixation, brain dissection, cryoprotection, cryosectioning, sequential RNA-FISH for cell-type identification, sequential DNA-FISH labeling, automated acquisition, and chromatin trace reconstruction. We also provide practical guidance for experimental design, sample preparation, tissue handling, image acquisition, and data analysis, highlighting critical steps that influence tissue integrity, hybridization efficiency, image registration, barcode detection, and chromatin trace reconstruction. The workflow is readily adaptable to different genomic loci and cell types, providing a robust approach for studying 3D genome organization in intact adult tissues.

Dimensional attention in learning is often implemented as a globally shared attention vector, where each stimulus dimension corresponds to a single scalar. These scalars are learned by models through gradient-descent on error, where predictive features acquire more salience. We show that under multi-outcome learning, where models predict more than one outcome, this shared vector becomes unstable; it collapses to its bounds and prevents the models from learning meaningful attentional tunings for learning and generalization. We address this by introducing an outcome-indexed attentional matrix that converts globally shared attentional tuning into an outcome-indexed representation. We present an analysis of the unstable shared vectors and derive the conditions under which it holds. Empirically, three synthetic experiments benchmark the proposed attention matrices and show that they converge to meaningful representations, something shared attention vectors fail to do. These results suggest that outcome-indexed attentional matrices are a general fix for gradient-based attentional processes, which improves models of learning under multi-outcome conditions.

Electroencephalography (EEG) microstate analysis commonly converts each scalp topography into a winner-take-all hard label and summarises the resulting sequence using duration, occurrence, coverage, transitions, and symbolic complexity. Although interpretable, this readout discards evidence strength, assignment ambiguity, and low-confidence periods. We introduce a template evidence trajectory framework that retains, at each sampled Global Field Power (GFP) peak, the evidence for all templates or subject-specific topographic communities. Conventional hard labels are treated as a compressed readout of this multivariate trajectory. We derive two evidence-aware extensions of classical descriptors: high-evidence episode duration and episode rate, which quantify temporal clustering and fragmentation of strong state evidence, and null-state Lempel-Ziv complexity (LZC), which explicitly encodes insufficient-evidence periods. We evaluated the framework across four resting-state EEG datasets spanning Alzheimer's disease and age-related variation. Fixed-state models included K-means, AAHC, and HMMs at K = 4 and K = 7, together with adaptive subject-specific Leiden and Infomap communities. Across 32 dataset-model cases, trajectory-derived duration effects exceeded matched hard-label effects in all cases at the representative setting and in 30-32 cases across a broader parameter grid. Null-state LZC improved over traditional LZC in most cases, while classification showed modest but consistent gains for trajectory or combined features. Retaining template evidence therefore provides a more sensitive readout of EEG topographic state dynamics while remaining compatible with conventional microstate analysis.

The brain uses discrete spikes for dynamic computation, yet, how neural microcircuits (NMCs) solve temporal credit assignment using local spike timing remains a fundamental open question. Dominant spiking neural network (SNN) approaches circumvent this by approximating backpropagation through surrogate gradients, decoupling learning from biological spike timing. Here, we reformulate temporal credit assignment as a state separation problem: extracting task-required components induced by historical perturbations directly from the current neural state. This enables an online feedback learning framework for NMCs through a gradient tunneling (GT) algorithm and the lead-lag expansion technique that derives credit assignment from local synaptic spike timing, while remaining compatible with ANN-SNN hybrid architectures. Experimentally, GT-trained NMCs excel at long-timescale evidence integration and noise-robust memory retention, and perform comparably to leading SNN online learning methods on real-world benchmarks with far fewer parameters. The proposed framework addresses the two-decade-old NMC feedback learning problem and suggests a computationally plausible explanation for the brain's learning mechanisms.

Electroencephalography (EEG) is a noninvasive technology that is widely used to monitor brain states, and many efforts are focused on developing reliable and efficient data analysis methods for EEG recordings. Here, we apply ordinal analysis to the EEG recording of the resting state of 109 healthy subjects measured in two different conditions: with eyes closed (EC state) or eyes open (EO state). We study the robustness of the temporal permutation entropy ($PE$) and the spatial permutation entropy ($SPE$) with respect to the presence of blinking artifacts in the EO recordings, the duration of the recordings, and the number of EEG channels analyzed. We perform a paired statistical test to assess whether these quantities differ significantly in the EO and EC recordings of the same subject. We find that $PE$ and $SPE$ perform surprisingly well, as they detect significant differences when calculated from the raw signals, even within a short time interval ($PE$) or from a reduced number of electrodes ($SPE$).

The quantum model of the two-sentence liar paradox proposed by Aerts, Broekaert, and Smets is an early example of the use of quantum formalism to describe cognitive dynamics. Our reconstruction is primarily pedagogical in intent, but it also leads to a number of clarifications, and to some new observations, concerning the structure of the model. Rewriting the model in Dirac notation, we make explicit the distinction between truth values originating from a decision and from semantic inference, and show that the associated enlargement of the one-sentence state space corresponds to a factorization with respect to which the non-paradoxical configurations are separable while the genuine liar state is entangled. We also derive the Hamiltonian generating the four-state cyclic evolution, express it in compact operator form, and give the resulting transition probabilities in closed form. We emphasize that the liar cycle admits a unitarily equivalent representation in the four-dimensional truth-only space, where the unmeasured liar state is separable, so the dimensional enlargement is not required by the unitary part of the dynamics; it is required, however, by the measurement structure. The enlargement is also required by the dynamics as soon as revision processes are admitted, in which states with identical truth content but different origins have different successors, the origin degree of freedom then acting as a minimal form of cognitive memory. We conclude by discussing the possible relationship between the model and more general deliberative processes.

We introduce a probability-wave framework for modeling the collective behavior of interacting adaptive agents, deriving testable eigenmodes through a generalized behavioral intelligence (GBI) nonlocal probability-wave equation. This framework captures a broad range of human intelligence behaviors with analytical mechanisms and offers an indirect method to examine the Liu-Chen-Ao (LCA) hypothesis of nonlocal entangled nerve fibers in the brain through collective trader behaviors. Our empirical analysis of Chinese intraday stock market data demonstrates that adaptive entangled game modes explain 82-94% (89% overall) of observed decision patterns, a sharp contrast to the predictions of neoclassical finance based on independent rational agents. Moreover, 2-12% of behaviors show adaption to intraday news, events, and environments, characterized by dual equilibrium states and abrupt reference point shifts, while purely independent modes occur in less than 5% of cases. These findings empirically support the LCA hypothesis, as observable trading behaviors reflect underlying brain mechanisms and internal intelligence decision-making in behavioral psychology. Our results highlight the necessity of incorporating adaptive entangled game modules into artificial general intelligence (AGI) architectures, addressing the limitations of conventional artificial neural network (ANN)-based AI, which relies on trillions of opaque parameters. By integrating ANN-based AI with probability-wave-based entangled-brain simulations, machine learning can enrich AGI foundation models (FMs) and facilitate the development of human-like processing units (HPUs) that leverage brain-inspired mechanisms. Such HPUs may ultimately create more compact, efficient, and robust AGI systems, particularly for embodied intelligence and robotics.

Motor control research involves the study of movement kinematics derived from the positional trajectories that complex motions describe. In natural, unconstrained motions requiring cognitive and memory processes in real time, the temporal speed profiles are not bell-shaped, may have multiple maxima and the peaks distribution is best fit by the continuous gamma family with two parameters, the shape and the scale. As the stochastic processes described by complex motion trajectories are non-stationary, the gamma shape and scale parameters describing them span stochastic trajectories. These points live on a curved surface where Euclidean distance depends on the arbitrary choice of parameterization. We adopt the Fisher-Rao distance (the geodesic length on the gamma manifold) as a coordinate-free metric for movement fluctuation signatures and present a robust numerical solver that converges across the full range of empirically observed parameters. We demonstrate the metric on a tablet-based digitized Trail Making Test in healthy adults, comparing micromovement fluctuations under low and high cognitive load. Load displaced every participant's fluctuation signature but with no shared direction. Instead, participants converged toward a common operating regime, with those farthest from it at baseline moving the most. Cognitive load thus contracts individuality in motor fluctuations rather than shifting the population uniformly.

Critical phenomena are widespread across many disciplines and have recently become a topic of deep interest in the study of biological and artificial neural networks. A distinct signature of criticality is the emergence of avalanches with power-law-distributed sizes and durations. However, empirically estimating the critical exponents remains challenging, and their interpretation is often model-dependent. In this work, we demonstrate how the Fisher Information Metric (FIM), a measure of generalized susceptibility, provides a comprehensive, model-agnostic characterization of the critical region in neural systems. We validate this approach across models of increasing biological complexity, from prototypical branching processes to spiking and whole-brain models, showing that FIM of each model's control parameter reliably tracks the system's degree of criticality. To emulate the study of real-world systems, where the control parameter is unknown, we additionally calculate FIM of the observed branching ratio of neural activity. The resulting FIM peaks where activity growth and decay balance, with the peak sharpening as the system approaches criticality. Hence, FIM peak width and height yield continuous, model-free readouts of proximity to criticality without requiring knowledge of the true control parameter, offering a robust tool for probing criticality in neural systems.

A common formalization of homeostasis is the free energy principle, a framework that defines a set of desired observation values, or critical states, that the agent should reach or remain close to. Under the free energy principle, an agent should act to maximize the probability of receiving the desired observations. Here we revisit the common approach of solving the problem of maximizing the log probability of the desired observations by maximizing a variational lower bound, the so-called negative free energy. We show that, instead, an approach directly maximizing that probability under the agent's policy is better suited to, and provides a better solution for, the original homeostatic control problem. This is done using hidden Markov model (HMM) control by allowing the policy to act over hidden states or noisy versions thereof while trying to maximize the probability of repeatedly having the desired observations. HMM control largely improves performance over the variational, or free energy, approach. We also show that the optimal policy is strictly deterministic, while the variational approach leads to a stochastic policy approximation. We finally provide a homeostatic reinterpretation of the maximum occupancy principle -a principle proposing that agents ought to maximize the occupancy of action-state path space -by defining homeostatic states as any states that do not immediately entail the termination or death of the agent.

Reservoir computing provides a controlled setting for studying how recurrent network architectureshapes computation: input signals are projected into a high-dimensional state space by a fixed nonlinear dynamical system, and only the readout is trained. However, reservoir performance can be dependent on hyperparameters; this paper asks which recurrent network features support robustness to those parameter changes. We characterize computational performance using memory capacity (MC), truncated single-delay information-processing capacity (IPC), and kernel rank (KR). Generalization across input histories is measured using generalization rank (GR), while hyperparameter robustness is quantified using the coefficient of variation (CV) of each metric across sweeps of target spectral radius, input scaling, leak rate, and neuron bias. To examine the architectural determinants of robustness, we construct perturbations that alter connectivity topology, excitatory/inhibitory sign structure, weight magnitudes, and weight placement while preserving complementary properties. Across these experiments, the C. elegans connectome consistently occupies a relatively low-variance regime. The central result is a performance-robustness tradeoff: architecture variants with higher task-agnostic performance also tend to exhibit greater hyperparameter sensitivity and poorer common-tail generalization. Across the E/I edge balance sweeps and shuffle controls, this tradeoff is closely associated with the raw spectral radius before normalization. Because every perturbed matrix is rescaled to the same target radius, matrices with lower raw spectral radius receive greater global amplification of their recurrent weights. The observed differences among architectures therefore characterize the joint effects of structural variation and architecture-specific global rescaling under spectral-radius normalization.

This paper examines the theory of Invariant Structural Learning (ISL), which proposes a non-optimization approach to concept formation. Learning is interpreted as convergence to structural attractors in a hypergraph space, rather than as the minimization of a global loss function. The paper presents the ISL model, including its mathematical formalization, computational verification, and a hypothetical neurobiological interpretation. The mathematical section introduces the formal apparatus of the structural reduction process and proves its finite convergence, the existence and uniqueness of class structural attractors, and the self-organization of attractor maps. The computational section demonstrates the feasibility of the proposed approach on classical image recognition tasks, utilizing the proposed learning mechanism without backpropagation and with extremely small training datasets. Finally, the neurobiological section formulates hypotheses regarding the possible implementation of structural attractors in dendritic trees, neural coding as a projection of internal attractor dynamics, and the development of neural architectures supporting the proposed learning concept. These hypotheses are discussed in the context of modern experimental data in the fields of dendritic computations, synaptic plasticity, and the structural organization of neural circuits. The proposed neurobiological mechanisms are presented as testable hypotheses rather than established biological facts. The results demonstrate the mathematical consistency and computational feasibility of the proposed model, while the neurobiological hypotheses outline potential directions for its experimental verification.