Papers

How can a rule inferred from examples reach someone who never saw them, without a shared code? Patients with severe aphasia do it by gesture or sketch. One network sees worked examples and emits eight invented symbols; a second, blind to them, applies them to a new input. Rewarded for the second's success, the first learns a code carrying rules to three-step transformations training never presents, which new learners acquire. Like invented human languages, the code has two regimes: under reward alone the speaker drifts to one message, as human languages lose words under plain transmission; expressive pressure keeps messages differentiated. Success on new rules tracks how much the message says about the rule, not how varied messages are. The learning signal shapes the code: reward sorts many rules under few fixed labels; the listener's error gradient gives each rule a region of similar messages, telling rules apart far better.

The firing fields of a grid cell form a hexagonal lattice, yet each field's intrinsic geometric structure remains unclear: Are individual grid fields radially symmetric or do they exhibit (weak) sixfold modulation inherited from the global lattice? To address this question, we quantified the within-field angular structure using harmonic analysis and per-cell matched simulations that preserved sampling statistics, field scale, and lattice geometry after correction for global elliptic deformation. We found that, across the investigated cells, the sixfold harmonic power fraction was consistent with values from matched locally circular reference fields. Experimental fields further showed substantially less sixfold modulation than matched simulations with an imposed local sixfold component, demonstrating sensitivity to sixfold structure at that level. This conclusion was supported by a complementary sixfold regression metric, harmonic leave-one-out sensitivity analyses, and robustness checks across radial-window choices, smoothing, and simulated background-noise levels. An observation-matched Burak–Fiete continuous-attractor network likewise showed no selective local sixfold enhancement despite its globally periodic grid structure, and this conclusion persisted when grid scale and relative field width were matched more closely to the experimental regime. These results suggest that the global hexagonal lattice organization need not be accompanied by detectable local sixfold angular modulation of individual firing fields, placing quantitative constraints on grid-field microstructure and mechanistic models of grid formation.

Understanding the neurobiological mechanisms underlying self-referential cognition and baseline self-consciousness remains a fundamental challenge in computational neuroscience. In this work, we propose a large-scale computational model incorporating a 10,000-neuron spiking neural network (SNN) based on Izhikevich dynamics. The network is structured into two interacting subsystems: a sensory processing layer (5,000 regular-spiking cortical neurons) and an endogenous Default Mode Network (DMN) pacemaker subsystem (5,000 intrinsically bursting neurons). The DMN layer is modulated by continuous tonic currents reflecting ascending brainstem neuromodulation, maintaining intrinsic, autonomous bioelectric rhythms independent of external sensory input. To represent top-down cognitive modulation, synaptic weights are hierarchically structured such that DMN-to-network projections exceed sensory-level connections. Through numerical simulations using a modified two-step Euler integration scheme, we demonstrate how endogenous pacemaker activity interacts with transient external sensory perturbations, providing an elementary mathematical framework for the emergence of a persistent, self-sustaining neural representation of "Self".

In neuropsychiatry, the primary goal is often not only to decode brain activity but to change it, for example to lessen a negative affective bias or an overly salient memory. Motivated by control theory, we develop an AI-driven neural-surrogate framework that proposes candidate representational changes and tests their predicted perceptual effects from snapshots of stimulus-evoked fMRI activity, without physical stimulation. The framework combines fMRI decoding, deep generative modeling, and constrained latent-space steering. Valence and memorability are used only as worked examples. Using more than 36,000 image-fMRI observations from four deeply sampled Natural Scenes Dataset participants, subject-specific models recovered coarse generative structure from visually responsive cortex (two-way identification, 0.79-0.88; chance, 0.5). Graded perturbations were reconstructed as images and evaluated with automated scorers and human ratings from 7,200 trials by 18 participants. In the primary VDVAE model, valence shifted from -0.61 to +1.03 SD and memorability from -1.34 to +1.45 SD; a later Versatile Diffusion refinement reduced or altered these effects. Across five perturbation levels, human valence ratings moved in the predicted direction under the linear time-correction model (mean slope, 0.038 SD per unit of alpha; 95 percent CI, 0.003-0.074; positive in 16 of 18 participants). Perceived memorability did not change reliably. Baseline agreement with the automated assessor was suggestive for valence (r = 0.30) and weak for memorability (r = 0.10). Extreme perturbations drifted from the original stimulus, so intended change must be weighed against loss of fidelity. These findings provide a falsifiable upstream method for designing and behaviorally testing candidate representational targets for future neuromodulation in psychiatry, while marking the limits of the present static approximation.

Biological sensory neurons have selective receptive fields organized along meaningful stimulus coordinates, such as frequency, motion direction, or retinotopic position. Such structure may arise from efficient coding and biological constraints on activity, connectivity, and wiring, as computational studies of simple neurons have shown across modalities. This raises a question: do structured receptive fields confer a computational advantage beyond resource efficiency itself, and does this advantage persist when individual neurons are highly expressive? We address this question in recurrent networks of Expressive Leaky Memory neurons, where we can independently vary neuronal complexity and the organization of feed-forward receptive fields. Across auditory and event-based visual classification tasks, receptive fields aligned with a task-relevant sensory coordinate improve test accuracy relative to budget-matched random receptive fields. This advantage disappears when sensory coordinates are scrambled, or when receptive fields follow task-irrelevant coordinates, showing that the benefit comes from alignment with task geometry rather than restricted connectivity alone. Increasing neuronal complexity reduces the performance advantage of structured receptive fields. Finally, generic synaptic sparsity regularization induces input selectivity and partially recovers performance, but remains substantially below explicitly structured receptive fields, suggesting that sparsity alone is insufficient to recover the full computational benefit of task-aligned receptive fields. Together, our results show that appropriate receptive fields can serve as a computational prior beyond sparsity itself, and that their value depends on the computational expressivity of individual neurons.

Infancy is a critical developmental window characterized by rapid functional brain reorganization, during which large-scale networks emerge, individualized connectome signatures continue to form, and early deviations may shape long-term cognitive and clinical outcomes. Functional MRI (fMRI) offers an opportunity to study these processes in vivo, yet extracting developmentally meaningful information from it remains challenging due to comparatively short scan duration, structured motion artifacts, variable scan states, and rapid brain maturation. Amid these challenges, deep learning has expanded the capacity of computational neuroimaging by learning robust representations from noisy, high-dimensional data, integrating complex spatial and temporal information, and capturing the nonlinear and rapidly evolving organization of the developing brain. Here, we review recent advances in deep learning for infant functional neuroimaging, synthesizing progress across input representation formatting, population and individualized brain mapping, longitudinal trajectory forecasting, robust and explainable model evaluation, and biological translation. Collectively, these methodological advances mark a paradigm shift in infant functional neuroimaging from descriptive, group-level analyses toward reliable, individualized, and developmentally grounded models. Future progress will depend on larger and more diverse longitudinal datasets, developmentally appropriate model designs, rigorous and standardized evaluation, and integration of computational predictions with biological mechanisms towards clinically meaningful outcomes. Addressing these priorities will help establish deep learning as a robust framework for understanding early functional brain development, identifying developmental variation at the individual level, and ultimately supporting earlier and precise assessment of neurodevelopmental risk.

Somatic membrane potential is the primary determinant of neuronal output, yet it remains inaccessible in many experimental setups where only dendritic recordings are available. Reconstructing somatic dynamics from distal measurements is a challenging inverse problem, particularly when the soma and dendrites are weakly coupled, as dendritic signals represent a filtered and attenuated version of somatic activity. To address this, we use a physics-informed neural network (PINN) constrained by a two-compartment Hodgkin–Huxley model. The network is trained on dense dendritic voltage recordings and the known injected current, complemented by a small number of somatic voltage samples (at most 5% of the time points), and it reconstructs the full somatic trajectory while adjusting a selected set of somatic maximal conductances. On synthetic data from four stimulation protocols, we quantify how the reconstruction depends on the amount of somatic supervision. Without somatic samples, the present formulation returns a smooth trajectory in which the action potentials are absent and the subthreshold level is biased, with a root-mean-square error of 10–14~mV; 1% of the somatic time points is enough to recover every spike; and 5% brings the root-mean-square error to about 2~mV, spikes included, with relative errors below 0.1% on the sodium and delayed-rectifier conductances. We also compare the PINN with an unscented Kalman filter constrained by the same model and assimilating the same observations, and we assess robustness to measurement noise and to random initialization. The reconstructions reported here therefore rely on sparse somatic anchoring in addition to the dendritic recordings. The results delimit what can be inferred in this synthetic, weakly supervised two-compartment setting and identify the amount of somatic information required by the present formulation.

Plasticity is invoked across the sciences to explain how systems can change, yet it is inferred from the very change it is meant to explain. A system may have many alternatives, realize none and still be plastic. Another may be driven far toward its only alternative, but the magnitude of that change does not establish its plasticity. What matters for plasticity is not how far the system moves but how strongly its present configuration constrains alternatives. Here I propose that plasticity, understood as a prospective property, is inverse configurational constraint: a system is plastic to the extent that its present configuration weakly constrains displacement toward alternatives within a declared representation. For systems sharing a functional partition, structural representation, and normalized architecture, increasing plasticity lowers every positive finite structural barrier separating the current state from its alternatives. With a fixed accessibility criterion, the accessible repertoire can expand but cannot shrink. In network representations, aggregate coupling is identified as configurational constraint, whose inverse defines plasticity. Barrier reduction and repertoire nesting are consequences, not components, of the definition. This connects two research traditions: quantifying connectivity without interpreting it as plasticity, and treating plasticity as a capacity for change without a prospective operational measure.

Psychophysics and decision research study perceptual binding and judgment contextuality apart. We argue both are scored against the same cyclic noncontextuality inequalities and share one convex global-consistency geometry – not one cohomology class – though only contextuality is tested, since binding's own residual vanishes here. Building on sheaf formulations of predictive coding (Seely 2025) and contextuality (Abramsky & Brandenburger 2011), a cyclic set of pairwise judgments admits a noncontextual explanation exactly when the cyclic (Suppes-Zanotti or n-cycle) inequalities hold, and its severity is measured by the complete contextual fraction CF (Abramsky, Barbosa & Mansfield 2017) rather than by the Cech invariant, which can miss it (Caru 2017). We build the perceptual arena from two binary judgments per cyclic-dominance pairing, scored against the same inequalities as the survey; an appendix rejects the one-bit alternative on construct-validity grounds. The central test is cross-domain: one cohort performs both arenas, and a shared latent tolerance predicts a positive correlation between their signed cyclic margins V*, the pre-clamp quantities behind CF. Both are inconsistency scores, so general response consistency confounds a bare correlation, and the prediction is therefore confound-residualized against a variance- and reliability-matched control. A positive result would support a shared residual association, not a common causal mechanism, which we prove this design cannot identify at any sample size. All three tests are designed but unrun, and the perceptual arena is a proposed instantiation with pilot gates – a methodological proposal, not a confirmatory report.

Pretrained EEG foundation models are increasingly proposed as general-purpose encoders for brain-computer interfaces, yet recent benchmarks disagree about when their representations transfer to downstream tasks. We audit LaBraM and CBraMod on motor imagery under a validation-locked protocol in which preprocessing, architecture, optimization, freeze depth, checkpoint, temperature, and method selection are determined using training-session data only. On four-class BCI Competition IV-2a, every supervised comparator evaluated here outperforms every foundation-model configuration, including validation-selected fine-tuning. We then examine a key confound: foundation models and task-specific decoders are normally evaluated using different input pipelines. Retraining three supervised architectures on the broadband arrays consumed by the foundation models produces matched-input accuracy differences of opposite sign across architectures: broadband input improves ATCNet by 0.078 accuracy while reducing EEG Conformer accuracy by 0.088. None of the three individual matched-input terms is significant after multiple-comparison correction at n = 9, so we treat the sign variation descriptively rather than as a formal architecture-by-pipeline interaction. These observed sign differences suggest that a single comparator may not provide an architecture-invariant decomposition of a pretrained-versus-supervised performance gap. The four-class deficit also does not reproduce uniformly across motor-imagery datasets: on two-class BNCI2014-004 we cannot detect the same separation between fine-tuned CBraMod and the supervised comparators. Finally, validation-fitted temperature scaling returns foundation-model calibration error to the supervised range despite substantially lower four-class accuracy.

Generative models of neural activity could help characterize tissue dynamics, compare experimental conditions, and simulate population activity for applications ranging from disease and drug-response studies to closed-loop experimentation. Existing approaches, however, typically assume a fixed set of sorted neurons, whereas high-density microelectrode arrays produce extremely sparse, array-wide binary spike volumes in which the observed subset of electrodes varies across assays. We introduce a discrete generative model that represents this activity using a shared vocabulary of spatiotemporal motifs. A residual vector-quantized autoencoder learns the motif vocabulary, while a factorized masked transformer predicts where activity occurs and which motif appears at each active location. We evaluate the model on 31 assays spanning human brain organoids and acute ex vivo human hippocampal tissue. The learned motifs are broadly reused: assay identity explains only $9%$ of the entropy in motif use, and motif overlap across tissue types is comparable to overlap within them. When representation quality is evaluated independently of the generative prior, our approach achieves $5.2\times$ the voxel-level reconstruction average precision of a matched flat tokenizer. For masked completion and free generation, the full model achieves $1.4$–$2.6\times$ the site-level average precision of the matched generative baseline and outperforms it across all four families of generation metrics. These results establish a compact, reusable representation for array-wide spiking activity without learned assay-specific parameters, providing a scalable foundation for generative modeling across diverse neural preparations.

This article asks under what conditions artificial intelligence could warrant a rational attribution of consciousness. Linguistic ability, multimodality, memory, planning, action control, and humanoid embodiment are not sufficient evidence of phenomenal experience. Biological precursors such as excitability, homeostasis, neural networks, and hierarchical representation instead identify functions whose counterparts may be engineered. The paper compares conventional LLMs, hybrid h-LLMs, vision-language-action systems, and embodied agents with the Motivated Emotional Mind (MEM) architecture. MEM adds receptor-grounded representations, regulatory self-monitoring and interoception, affect, semblion-based associative memory, and re-entry into lower sensory and interoceptive maps. A related formulation defines a phenomenal state as a dynamic, integrated state of the whole embodied system in which sensor-grounded modal content is recurrently stabilized and coupled to a suprathreshold interoceptive representation of the system's regulatory state. The mathematical model of MEM published on arXiv specifies implementable relations among these components and makes the proposal more precise and testable. Comparative and causal tests in humans, invertebrates, and artificial systems could strengthen or weaken the MEM capabilities. MEM is therefore presented as a falsifiable research program requiring empirical validation and ethical caution.

Epileptic seizures arise from complex, nonlinear interactions within brain networks, yet reliable electroencephalographic (EEG) prediction remains challenging due to the nonstationary and heterogeneous nature of neural dynamics. Existing methods typically analyze EEG data as static or weakly time-dependent snapshots, overlooking the intrinsic dynamics and lacking the geometric sensitivity to capture the hierarchical, localized evolution of the epileptogenic zone. To address these limitations, we propose an offline, patient-specific evaluation of chaotic dynamics-regulated topological learning (CDRTL) for distinguishing preictal from interictal EEG states. This framework unifies chaotic dynamics, multiscale algebraic topology, and local network differentiation. Specifically, we partition EEG signals into discrete functional subnets based on correlation strengths, capturing the multi-scale connectivity of the brain. By modeling each node as a Lorenz oscillator, we embed the underlying chaotic dynamics into the network architecture. We then apply the persistent Laplacian to simultaneously extract topological invariants and geometric shape evolution through harmonic and non-harmonic spectral analysis. Additionally, a node-removal topological differentiation strategy isolates localized neural contributions. Our framework was evaluated on the CHB-MIT database using balanced preictal and interictal labels and stratified channel-level cross-validation within each patient. The results support offline discrimination of preictal and interictal channel-level nodes within fixed patient-specific networks. Because representations are constructed from the complete network, including held-out unlabeled nodes, before cross-validation, the reported performance is specific to this transductive setting and does not establish generalization to unseen EEG windows, seizures, or patients.

Advances in large-scale neural recording have made it possible to collect data across many animals and distributed brain regions, raising the question of whether this scale can be exploited to learn general-purpose neural representations transferable across diverse downstream tasks. Yet, progress toward this goal has been limited by fragmented evaluation protocols and a narrow focus on individual task domains. Here, we present BrainWideBench, a benchmark for evaluating across-animal transfer on multi-region neural recordings, built on the International Brain Laboratory Brainwide Map dataset of neural and behavioral recordings spanning 276 brain regions from 139 mice performing a sensory-guided decision-making task. The benchmark is organized around three complementary task suites that evaluate whether learned representations support downstream decoding of behavior, can predict masked or future neural activity, and can recover biologically meaningful anatomical organization. With this benchmark, we systematically evaluate pretraining methods across transfer settings, including finetuning on downstream objectives and zero-shot generalization to unseen animals. Our results confirm pretraining improves performance over matched single-session baselines, but we show current methods exhibit heterogeneity in transfer capabilities: gains depend strongly on the alignment between pretraining objectives and downstream tasks. No single approach performs uniformly well across all three suites, and most methods are designed to only address a subset of them. Together, these findings suggest that learning representations that jointly generalize across behavior, dynamics, and anatomy remains an open challenge. By providing a unified and reproducible evaluation suite, BrainWideBench establishes a framework for measuring progress toward general-purpose models of the mouse brain.

Memory segmentation is thought to arise from rapid decorrelation in neural activity, often quantified by Euclidean distance or cosine angle. Although these metrics detect a transition, they do not reveal how the new state relates to the repertoire represented by the neural manifold. This matters because neuromodulators that drive state transitions also alter excitability, and learning may repurpose existing representations or create new ones. Here, I introduce a geometric decomposition that separates changes attributable to gain modulation of a nearby manifold state from movement within the manifold and genuine off-manifold displacement. The approach uses the radial axis of neural population activity to partition the normal space of a local manifold region. A central challenge is identifiability: given only a static reference manifold and a single test state, neither the state from which a perturbation began nor its gain magnitude and mechanistic decomposition can generally be recovered uniquely. I therefore formulate identifiability as a cascade of geometric gates specifying when each component can be interpreted. The gates distinguish structural failures, including the absence of a local chart or incorrect intrinsic dimensionality, from estimation error and systematic bias caused by reference sampling, tangent-frame error, gain-axis misalignment, anchor displacement, and poor ratio conditioning. Simulations show that neighborhood size, curvature, sampling density, ambient dimension, and noise act through a small set of geometric quantities. The framework specifies when assignments to gain or novelty are identifiable, how they become biased, and which diagnostics reveal the relevant failure regime. By quantifying the nature rather than only the magnitude of neural state change, it provides a clear, readily interpretable framework for evaluating mechanisms of neural state transitions.

Dendritic silhouettes are cheap. Pond ice neurons, diffusion-limited aggregates, wiring-minimizing arbors, and biological neurons can share a branching statistic because they share a growth instability, not because they are functionally similar. We take that observation as a modeling constraint. The ice neuron only structurally resembles a biological neuron. A biological neuron carries an internal state and a field that it helps generate, letting geometry, dynamics, and field co-sculpt one another. In our construction, the ice neuron is the zero-gain limit of an interfacial system in which a FitzHugh–Nagumo state lives on the moving front and writes back into the field that drives it, with dimensionless gain $\gamma$. At $\gamma=0$, morphology and internal dynamics decouple into a Mullins–Sekerka growth band and a damped oscillator. Finite $\gamma$ deforms the spectrum, shifts the selected instability, and opens an oscillatory region below the isolated Hopf threshold. On a dendrite grown at $\gamma=0$ and then frozen, field closure increases distal persistence without changing the geometry; its first-order effect remains positive across repeated realizations of the same growth rule and localizes to a small set of non-cable field shortcuts. Nonlinear dynamics remain finite beyond the linear instability: below threshold, closure is almost the transmission; at spike amplitude, it is a correction to a regenerative cable event. A dendritic tree therefore is not intrinsically a structural spandrel. Shape alone is morphology without an internal state or a source term. That is the dendritic silhouette of the ice neuron. Excitability adds an active cable. Field closure is the biophysical mélange in which the neuron helps write the field that acts back on it, sculpting its dynamics and, during growth, its structure. That is a biological neuron.

Movement disorders (MDs) frequently co-occur, yet phenomenological and severity assessment shows substantial inter-rater variability. Markerless video could improve reproducibility, but prior work is largely single-symptom, depends on standardized acquisition, and lacks validation and transfer across ages and sites. We combined two foundation models into one frozen backbone: Segment Anything Model 3 (SAM 3) for dense, per-frame markerless segmentation summarized into geometric, contour and grid kinematic signals, and TabICLv2, a tabular foundation model, for in-context multi-label classification of eight hyperkinetic MD phenomenologies. Trained on standardized recordings of 21 adults and 4 controls, it transferred unchanged to two independent datasets, pediatric (n=12) and tremor-dominant adult (n=20), assessed with the CODY-SAMP scale; only the patient-level decision step was recalibrated per site. Under clinician consensus labels, false positives fell to zero in both datasets. Dystonia recovered perfectly (7/7 pediatric; 15/15 adult held-out), chorea fully in children (3/3), and tremor was recovered in adults (11/15) once a tremor-rich cohort made it evaluable, through recalibration alone. Per-region effect-size analysis gave clinically coherent, phenomenology-specific signals and identified myoclonus as the principal failure. Against YOLOv8 sparse keypoints, the dense representation matched under clinician permissive labels (Jaccard 0.63 vs 0.63) and was markedly more robust under clinician-label consensus (0.93 vs 0.76). This frozen foundation-model backbone with light per-site calibration yields transferable, interpretable, conservative multi-label phenotyping of co-occurring hyperkinetic MDs across ages and from standardized to routine video, adding robustness on high-confidence, clinician-agreed labels. Prospective multi-centre validation is required before clinical use.

Computational models of auditory physiology commonly target specific responses or stages of the auditory pathway, limiting their ability to integrate findings across experimental paradigms and neural timescales. We present a foundation model of human auditory electrophysiology: a causal neural network trained to map binaural acoustic waveforms directly to high-sample-rate EEG. The model was trained on approximately 250 hours of EEG data from 92 subjects, with varied electrode montages and stimuli spanning tonebursts, speech, and music. We tested whether the model recovered effects of stimulus rate, frequency, and presentation method on auditory brainstem responses (ABRs); subcortical and cortical temporal response functions (TRFs) to continuous speech; and the click-evoked binaural interaction component (BIC). Predicted ABRs and TRFs reproduced established response morphology and stimulus-dependent effects, with model-grand-average correlations falling within the corresponding subject-level human distributions. The model-predicted BIC metrics closely resembled the values reported in the literature. These findings demonstrate that a single audio-to-EEG model can capture auditory physiology across paradigms and timescales, supporting future in silico experimentation and hearing technology applications.

This article presents a mathematical model of the Motivated Emotional Mind cognitive architecture developed for embodied intelligent systems. Such a system learns to maintain its homeostasis through a generalized form of reinforcement learning based on its internal motivations, termed motivated learning (ML). The principal contribution of this article is a rigorous formalization of the re-entrant loop integrating feedforward processing, lateral interactions, and feedback pathways, together with the representational selection mechanisms that govern adaptive system responses. The model specifies how ongoing exteroceptive and interoceptive signals, bodily-motivational context, and memory traces are bound into associative memory structures termed semblions, which compete for access to further processing and top-down reconstruction. The formalization encompasses secondary perception, representational competition, curiosity, procedural gaps, and action selection directed toward limiting allostatic violations. Within this framework, motivated learning is tailored to embodied systems whose dynamics are shaped by needs, affect, and the current regulatory state. Unlike standard reinforcement-learning models, the proposed approach incorporates need thresholds, goal generation and shifting goals, bodily state, resource constraints, and action uncertainty, thereby providing a more adequate account of response selection under regulatory pressure. Global affect functions as a central control signal, modulating the learning rate, representational valence, and the balance between exploration and exploitation. The model presented here is a step toward a more rigorous formalization of cognitive phenomena and may provide a basis for further theoretical analysis, computer simulation, and implementation in artificial-intelligence systems inspired by biological processes.

The brain needs an accurate internal model of the world to generate predictions and guide behavior. However, it must estimate the statistical structure of the environment from limited experience. This is particularly difficult for rare events, whose observed frequencies in a limited sample may substantially under- or overestimate their true frequencies. How the brain constructs an accurate internal model despite this sampling problem remains unclear. We address this problem using a Bayesian Confidence Propagation Neural Network (BCPNN) trained on event sequences from a Markov-chain random walk with controlled event frequencies. Treating the underlying Markov structure as the ground truth, we train the network on limited sample of event sequences and then allow it to generate autonomous replay based on the learned structure. We evaluate replay fidelity at the levels of both marginal event frequencies and conditional transition structure. We find that moderate neural noise, modeled as temporally correlated random fluctuations in unit activity during replay, is critical for faithful internal simulation. Without this variability, deterministic replay systematically under- or overrepresents rare events, whereas moderate noise restores both their marginal and conditional occurrence. Moderate noise also broadens the range of parameter values that produce accurate replay, making the model more robust to parameter variation. Together, these results support noise-assisted internal simulation as a potential mechanism for compensating for sampling errors arising from limited experience. Our model also provides a testable framework for investigating how altered neural variability may impair internal-model fidelity in disorders such as Parkinson's disease.

The free-energy principle (FEP) and active inference describe adaptive systems through probabilistic generative models and free-energy functionals, whereas integrated information theory (IIT) characterizes intrinsic cause-effect organization. We separate two questions: whether revised IIT quantities admit local free-energy-like representations and whether scalar expected free energy can determine system integrated information. We show that the 2026 IIT intrinsic-differentiation and specification terms admit exact surprisal representations, and hence variational-free-energy representations under exact inference. We then construct noisy binary permutation networks whose causal topology varies while local transition reliability is fixed. In a homogeneous state, all permutations yield the same one-step predictive distribution and state-risk expected free energy under common preferences and observation models, yet revised IIT distinguishes them: multicycle permutations have zero whole-system integration, whereas a single noisy cycle has positive integration at intermediate reliability. For an n-unit noisy single cycle, we derive a closed-form expression for revised system integrated information, prove that the minimum-information partition cuts exactly two causal edges, and show that as n grows the optimal reliability approaches unity, the joint specified-state probability approaches one half, and the maximum integration approaches one ibit. Numerical IIT calculations for n = 2-5 and exhaustive partition enumeration through n = 8 reproduce the analytical results. Thus, local surprisal/free-energy correspondence coexists with global non-equivalence: equal expected free energy does not imply equal revised IIT system integrated information. Any synthesis of IIT with FEP/active inference must retain counterfactual causal information beyond a scalar free-energy value.

Training spike detectors on real recordings is challenging because algorithmically generated labels can be noisy and incomplete. We propose VanillaSort, combining multichannel detection with spatially augmented, template-guided clustering. VanillaDet uses visibility-aware masking, truncated Gaussian targets and a temporally tolerant positive-bag loss, followed by conditional event-SNR gating. VanillaCluster combines HuiduRep embeddings with relative-amplitude features for Gaussian mixture clustering and refines assignments using cross-fitted waveform templates built from selected core events. VanillaDet improves detection accuracy over SimSort by two and three percentage points on the static and drift subsets of Hybrid Janelia, respectively. The complete pipeline also improves sorting performance over the corresponding HuiduRep baselines. These results support learning from imperfect real-data labels and incorporating waveform consistency into neuronal assignment.

Learning flexible motor primitives is a hallmark of skilled motor control. Recent neuroscience theory proposes that motor primitives may be implemented as low-rank perturbations of a shared recurrent network, but leaves open how such a system is learned. We translate this principle into a novel architecture for learning motor skills end-to-end: a shared recurrent core modulated by a bank of residual adapters, each selected by a discrete latent code. Trained on closed-loop biomechanical control, the adapters develop emergent low-rank perturbations of the recurrent dynamics despite no architectural rank constraint, placing task representations in disparate subspaces of the shared core network. A simple high-level policy over the learned options, optimized while the whole network is frozen, sequences the low-rank adapters to produce novel out-of-distribution movements. We demonstrate the ability to generalize to novel motor sequences within the closed-loop control setting, improving on the generalization error of a task-input-conditioned multitask baseline by upto order of magnitude.

Spaced repetition software schedules reviews with a memory model fit to review logs. Accuracy on a test set is not sufficient evidence of quality since available data are produced by existing schedulers, and new solutions must extrapolate beyond them. A model also needs a simple mechanistic interpretation. We present SBD, a model that is more interpretable and 80% smaller than the current state of the art at nearly the same accuracy.

Testing an artificial system for a property linked to consciousness means applying a measurement developed on brains to a system that is not one. Such a transfer must re-examine five parts of the procedure: the competing statistical models, how they are fitted, the unit the inference generalizes over, the quantity the uncertainty interval is about, and the rule that turns a result into a verdict. The Cross-Substrate Access Assay declares all five. Because brain and model signals share no physical scale, every model is scored by the cross-entropy it assigns to held-out data, in nats per trial. The test case is the global neuronal workspace theory, which predicts that near threshold a stimulus either enters a capacity-limited workspace or does not, so that single-trial responses form a mixture of two states. A published test of this prediction on twenty people's electroencephalograms partly reproduces in a re-implementation: the first crossing and the broad ordering over time match, the window-by-window agreement does not. On 12,000 synthetic datasets generated with a single graded state, all of them members of the families the procedure fits and none within 0.0067 nat per trial of the decision boundary, the two models of that test carried over unchanged reported two states in 989 and the expanded families in none; on 600 datasets carrying a mixture the expanded procedure reported two states in 599. Its nominal 95% interval contained the procedure's mean result less often than the required 90% at six of twelve graded settings. No claim about experience is made.

Foundation-model features are increasingly used to ask what information neural activity represents, often by comparing prediction gains between nested encoding models. We show that such multimodal contrasts can change sign when only the conditioning predictor is reconstructed. Using fMRI from the Natural Scenes Dataset, DINOv2 visual features, and MPNet embeddings of MS COCO captions and Localized Narratives, a caption-narrative contrast in the additional predictive contribution of vision favors narratives when one short caption is compared with a long narrative (+0.012/+0.015 in Places), but favors captions after approximate word-count matching (-0.031/-0.023). The shift occurs across every measured ROI in both subjects and is driven primarily by differences in language-only prediction. Comparable contrasts also survive removal of image-specific content-word identity in several ROIs. These results show that nested neural contrasts do not identify represented content by themselves: predictor construction is part of the experimental design, and matched controls are required for representational claims.

Astrocytes are non-neuronal glial cells that are receiving widespread attention due to their emerging role in neural computation. In this paper, we propose and study dynamical mechanisms by which astrocytes may augment the ability of neural networks to infer context in reinforcement learning (RL) settings. We construct a biologically inspired, two-level dynamical neural-astrocyte network with distinct spatial and temporal organization. We train this model on a hierarchical multi-context task that requires the agent to infer changes in latent task rules based on derived rewards. We find that in this setting, astrocytes enable evidence accumulation of changes in context and subsequent context-specific modulation of neural dynamics. We show that these functions are implemented via two dynamical mechanisms: (i) reward-induced bifurcations that relocate an asymptotically stable attractor into different, context-specific regions of state space, and (ii) the relative shallowness of these attractors, mediated by the entropy of the environment, giving rise to behavioral stickiness. Together, these mechanisms amount to a hybrid automaton, in which uncertainty accumulates until, eventually, the neural dynamics are switched to a new context. This model provides a neuro-dynamic schema, compatible with neural-astrocyte biology and prior empirical observations, for how astrocytes may integrate information from the periphery and drive contextual changes in neural circuits.

Repeated decisions exhibit temporal structures such as persistence, direction-dependent switching, recurrent alternation, and abrupt transitions. We examine the generative sufficiency of a two-dimensional fast-slow dynamical system. The system combines a cubic fast equation with linear slow feedback and is analyzed through its equilibrium geometry, trace-determinant structure, equilibrium-fold loci, candidate Hopf boundaries, and singular critical manifold. An explicit observation operator projects continuous trajectories to a scalar signal and applies a binary readout, separating latent state-space dynamics from observable behavior. The analysis establishes a unique-equilibrium regime and, for suitable parameters, a three-equilibrium wedge with a central saddle. The outer equilibria are attracting only where their traces are negative. The analysis also identifies the simple-zero condition required for ordinary saddle-nodes, trace-zero positive-determinant spectral boundaries compatible with oscillatory instability, and the attracting and repelling branches of the critical manifold. Prescribed nonautonomous sweeps numerically illustrate direction-dependent switching (T1), transient episodic recurrent switching (T2), and an abrupt localized regime shift (T3). The attracting-equilibrium regime associated with prolonged state retention (T4) is characterized analytically. The resulting correspondence is intended as a test of generative sufficiency at the level of observable temporal organization, rather than as an identification or empirical validation of latent cognitive mechanisms.

Achievement emotions shape how students engage with and learn from academic tasks, yet most studies examine individual emotions rather than co-occurring affective profiles and their dynamics. We examined latent achievement-emotion profiles and their transitions following exposure to different instructor facial expressions during a video lecture. Self-reported data from 78 Grade VII and VIII students revealed three profiles: enthusiastic, demotivated, and vulnerable. Profile transitions differed across instructor conditions, with happy expressions favouring more adaptive transitions and angry expressions favouring transitions toward demotivation. Exploratory factor analysis and Bayesian structural modelling further identified preparedness and cognitive restraint as regulatory dimensions associated with profile switching.

Classical theories of random neural networks typically assume independent connectivity, overlooking the local motif structures prevalent in biological circuits. Here, we investigate how four second-order synaptic motifs (chain, reciprocal, convergent, and divergent) shape the dynamics of nonlinear firing-rate networks. While previous studies have established that chain correlations generate outlier eigenvalues, we demonstrate that these motifs also jointly reshape the Jacobian eigenvalue bulk. Using the path-integral formalism, we derive a dynamic mean-field theory which reveals that the chain motif acts as a retarded feedback of the ensemble-mean activity through the response kernel, producing a rich repertoire of dynamical regimes, including ferromagnetic states and limit cycles. At sufficiently large magnitude, negative chain correlations produce a glassy, multistable regime that was previously mainly associated with partially symmetric networks. Our theory also distinguishes convergent from divergent motifs: divergent correlations primarily rescale temporal noise, while convergent correlations suppress temporal chaos by converting nonzero mean activity into quenched heterogeneity. Finally, analyses of the Lyapunov spectrum and participation-ratio dimension show that motif structure changes the geometry of chaotic activity, reducing entropy production and attractor dimensionality even when the effective spectral edge is held fixed. Together, these findings establish second-order motifs as a fundamental structural mechanism governing the dynamical regimes of local cortical circuits.