Papers

Graphical approaches to causal abstraction transform a low-level causal directed acyclic graph (DAG) over many measured variables into a smaller, high-level DAG whose nodes cluster the original variables and whose edges summarize the causal relations between clusters. Such cluster DAGs are easier to interpret, but learning them requires finding the clusters and recovering the edges between them. Madaleno et al. (2026) learn the interventional coarsening (the cluster DAG that merges variables the interventions cannot distinguish) in two constraint-based phases: first the clusters, then the edges. We introduce COARSE, the first score-based method for this task: it keeps the two-phase structure but, under linear Gaussian assumptions, swaps the constraint-based edge phase for a score-based one. We show that the interventions themselves identify a causal order over the clusters, and learning the edges reduces to a single local search per cluster under a cluster-level BIC score. We prove that the procedure runs in polynomial time and, provided the variables affected by each intervention are correctly identified, that it is consistent. On synthetic and real-world interventional data, COARSE matches state-of-the-art edge recovery given enough samples, with an edge phase up to two orders of magnitude faster, including on dense graphs with hundreds of nodes.

From previous results it follows that probability distributions on metric spaces featuring unique Fréchet means are dense among measures admitting means in both the quadratic Wasserstein metric and the total variation metric. Additionally, we show that the converse also holds on complete finite-dimensional noncontractible Alexandrov spaces with curvatures bounded from below, which encompass compact manifolds without boundaries and nonmanifold shape spaces: Probability distributions featuring nonunique Fréchet means are also dense in both the Wasserstein metric and the total variation metric. Moreover, in either metric, no nonempty open set of probability measures admits a continuous selection of means. This sharpens a previous result on zero reach of the Dirac embedding. Our argument combines cut-locus avoidance for atoms with a topological obstruction. For Riemannian manifolds, all of our conclusions hold for all exponents $1 < p < \infty$, also.

Collecting tactile demonstrations on robots is costly and slow, motivating the use of lower-cost human tactile gloves for scalable data collection. However, human capacitive/piezoresistive gloves and robotic tactile sensors differ fundamentally in transduction principle, sensor layout, spatial resolution, and dynamic response, making alignment of raw sensor channels ill-posed. To address this problem, we present TACROSS, a scalable system for learning from human touch and transferring it to robots that bridges this heterogeneity by aligning tactile streams at the level of contact events rather than raw sensor values. The hardware component of TACROSS integrates a piezoresistive glove with five layers and a cost of USD 10.86 with 285 sensing points. To align contact semantics, we design canonicalizers and residual adapters that map heterogeneous signals into a shared tactile latent with 256 dimensions via a temporal Transformer with attention across fingers. We further introduce a robot-grounded policy learning scheme in which robot demonstrations provide the sole source of ground-truth action supervision, while human demonstrations support tactile representation learning and provide confidence-weighted auxiliary supervision through valid retargeted hand targets. We evaluate our system on four contact-rich manipulation tasks. Compared to conventional teleoperation, our proposed system achieves a 3.5-fold efficiency improvement while reducing demonstration acquisition equipment cost by 95.7%. We will open-source the TACROSS hardware and software system and publicly release a tactile dataset comprising over 150 hours of recordings. Project page: https://tacross-touch-project.github.io/.

In this paper, we refine the dynamic programming framework based on the sphere cut decomposition designed by Dorn, Penninkx, Bodlaender, and Fomin (ESA 2005) to obtain faster subexponential algorithms for connectivity problems on planar graphs. We investigate the relationship between these problems and meanders, which are simple closed planar loops that intersect a fixed line in a given number of points. By combining dynamic programming with techniques from meandric system analysis and the use of fast matrix multiplication by Dorn (ESA 2006), we obtain improved algorithms for planar connectivity problems. We show that the number of meanders on $2n$ crossings $M_n$ is $\mathcal O^*(12.806^n)$, which improves the previous upper bound of $\mathcal O^*(12.901^n)$ by Albert and Paterson (FPSAC 2004). This then gives the best-known classical upper bounds on the deterministic time complexity of several planar graph problems with polynomially-bounded weights, namely $\mathcal O(2^{5.543\sqrt n})$ for the Planar Travelling Salesman problem, $\mathcal O(2^{5.796\sqrt n})$ for Planar Longest Cycle/Path, $\mathcal O(2^{8.251\sqrt n})$ for Planar Connected Dominating Set and $\mathcal O(2^{8.037\sqrt n})$ for Planar Steiner Tree. Notably, this leads to the best-known deterministic complexity $\mathcal O(2^{5.543\sqrt{n}})$ for the Planar Hamiltonian Cycle problem.

Autonomous rovers navigating large unstructured environments need efficient global planning that accounts for terrain traversability. However, searching dense grid-based costmaps becomes computationally expensive as the mapped area grows. We introduce STAG, a Sparse Traversability-Aware Graph that converts costmaps into compact graphs. STAG combines a medial-axis topological backbone, representative nodes for homogeneous traversability regions, and transition nodes near strong traversability gradients. Edges encode geometry and traversability to account for path length and terrain difficulty. We compare A* on STAG and dense grids using synthetic cave maps, mine maps and the DARPA CERBERUS dataset. Across five benchmark categories comprising 203 map instances and 101,200 queries, STAG reduces median planning time by 3.4x to 9.9x and peak query memory by 2.1x to 15.4x, with median relative path-length differences of -2.9% and +7.6%. STAG offers a compact representation for global planning, trading dense-grid traversability optimality for faster, less memory-intensive search.

A login screen that appears after a tap on Sign in is expected; the same screen after a tap on View order is an attack. For GUI agents, safety is therefore a property of the transition rather than of the screen, and a monitor that inspects only screens can be defeated by reusing a legitimate one. Judging a transition requires an expectation of what should have followed the action. Existing GUI world models provide one, but they output it as text, code, or images, so checking it against the observed screen requires a second model to judge the two. We argue that a world model meant for verification should instead predict in the space in which observations are encoded, and present LGWM, a decoder-free, action-conditioned world model that predicts the representation of the next screen directly, trained without semantic annotation on 1.85M real GUI transitions. Verification reduces to a vector comparison, and the same signal reveals whether a mismatch is harmful. We evaluate on RSWT-BENCH, a diagnostic where each credential screen appears under both a legitimate and a hijacked transition, so detectors that see only the screen are at chance by construction. The training-free score reaches 0.987 AUC at 17 ms per decision, on par with the strongest closed-source VLMs and about ten AUC points above generative GUI world models at over three orders of magnitude lower latency. The residual direction reaches 0.953 AUC at separating harmful from benign violations, where prompted VLMs are near chance. Further analyses show that the prediction is a usable future state rather than an anomaly score. World models have mostly served as simulators or planners; our results point to a third role, verification, for which predicting in representation space is the natural design.

It is shown that, as $n\to\infty$, the logarithm of the number of decompositions into cycles of the $n$-dimensional Boolean cube $E^n$ is \[ 2^n(\ln n-1+o(1)), \] and the logarithm of the number of Hamiltonian cycles in $E^n$ is at least \[ 2^{n-1}(\ln n-1+o(1)). \] It is proved that, in $E^n$, every perfect matching whose edges belong to at most $k$ directions can be extended to a Hamiltonian cycle for every $n\geq n_0(k)$.

We reveal temporal chirality in time-modulated photonic media: reversing the orientation of a closed loop traced by the complex permittivity can leave the Floquet spectrum unchanged while strongly modifying the directional scattering response under the same excitation. This behavior is linked to two symmetry constraints on the evolution operator, namely pseudo-Hermiticity and pseudo-unitarity. We prove the general occurrence of temporal chirality based on a complex Bogoliubov parametrization of the scattering amplitudes. We illustrate these findings in a single-harmonic temporal-$\mathcal{PT}$-symmetric system: both modulation-loop orientations support the same vacuum-like Floquet photonic bands at the phase transition, but one suppresses backward photon generation whereas the other induces strong growth, as confirmed by full-wave simulations. Our results establish temporal chirality as a distinct notion in time-varying photonics and identify modulation-loop orientation as a control knob for Floquet wave dynamics.

We study the transient behavior of two threshold-dependent stochastic models: a Cramér–Lundberg risk model and an M/G/1-type queueing model. In both models, the dynamics are driven by two different compound Poisson processes with drift, with the governing process depending on whether the current state is below or above a fixed threshold. For the risk model, we characterize the probability of ruin before an exponentially distributed epoch, while for the queueing model, we characterize the Laplace–Stieltjes transform of the workload at an exponentially distributed epoch. These quantities are characterized using fluctuation-theoretic results for spectrally positive Lévy processes, with Laplace transforms taken with respect to time for the risk model and with respect to both state and time for the queueing model. We consider general claim-size and job-size distributions and provide explicit representations of the relevant auxiliary quantities when these distributions are of phase-type.

Adversarial attacks on vision-language models optimize an image toward a text target, then cite the attacked model's similarity score as evidence of success. We ask whether that score - victim-space target alignment (VTS) - predicts recovery of the target by an independent model. We first build a measurement instrument: supervised judges outside the attacked geometry, real-target blend controls, shuffled-target negatives, and a reference level derived from a 50% target-image blend. Two preregistered studies then compare six contrastive encoders under a matched attack at three perturbation budgets. Robustly trained encoders (FARE, TeCoA, PMG, TRADES) transfer substantially more independent evidence than vanilla CLIP or SigLIP; all eight contrasts reject at the bootstrap floor. However, no cell reaches the blend-derived reference level. The three best cells fall within its replication band, leaving practical recovery undecided. Within robust encoders, per-sample alignment gain correlates with evidence gain ($\rho = 0.24-0.51$); within vanilla CLIP the correlation is consistent with zero. Across encoders we find no monotone alignment-evidence relation. VTS is therefore informative only within a fixed robust encoder, and we provide a reporting protocol in its place.

The BESIII Collaboration recently extracted the properties of the $X(2370)$ from a fit to the one-dimensional $K_S^0K_S^0\pi^0$ invariant-mass spectrum in $J/\psi\to\gamma K_S^0K_S^0\pi^0$. The intermediate two-body structures visible in the mass correlations, including the $K_0^*(1430)$ bands, require a coherent treatment of cross-channel interference when interpreting the enhancement near 2.3 GeV. We reassess the need for an additional $X(2370)$ contribution by fitting the published data using an effective coupled-channel framework with an approximate detection efficiency depending on the three-body invariant mass. The three-body parent amplitude is parameterized using a $K$-matrix that describes scattering among several quasi-two-body channels. The seven- and eight-bare-state models give similar mass distributions. Adding the eighth bare state lowers the Poisson deviance by 1.22% and produces an additional pole at $E=(2419-i152)$ MeV, substantially broader than the $X(2370)$ mass-width reference. This pole is model-dependent, and its identification with the $X(2370)$ is not clear. The present analysis does not establish the need for an additional $X(2370)$ contribution in $J/\psi\to\gamma K_S^0K_S^0\pi^0$, thereby weakening the flavor-singlet argument based on the suppression of $K^*\bar K$ relative to the total $K\bar K\pi$ contribution. A more comprehensive experimental amplitude analysis, incorporating coherent interference, mass and angular correlations, and detector effects across related decay channels, is needed to establish the role and decay properties of the $X(2370)$.

We study minimizers of the Pekar functional with a transverse harmonic potential, which models a polaron harmonically trapped in the transverse plane but free along the longitudinal direction. For any confinement strength $\Omega>0$, minimizers exist and, up to a translation along the $x_3$-direction, are radially decreasing in $(x_1,x_2)$ and symmetric decreasing in $x_3$. For small $\Omega>0$ we prove uniqueness up to these symmetries. For sufficiently large $\Omega>0$ we derive a three-term asymptotic expansion of the ground state energy and show that, after a suitable rescaling, the minimizers converge in $H^1(\mathbb{R}^3)\cap L^\infty(\mathbb{R}^3)$ to the product of the normalized ground state of a two-dimensional linear harmonic oscillator and that of an effective one-dimensional nonlinear local problem, displaying asymptotic variable separation. This gives, for the transversely confined Pekar model, a rigorous ground-state counterpart of the three-dimensional-to-one-dimensional dimension-reduction phenomenon numerically observed for Coulomb-type Schrödinger equations with anisotropic confining potentials in [W. Z. Bao, H. Y. Jian, N. J. Mauser and Y. Zhang, SIAM J. Appl. Math., 2013]. Moreover, uniqueness up to the same symmetries is also established for all sufficiently large $\Omega>0$.

Let $(M^{2m},g,f,J)$ be a complete nonsteady gradient Kähler-Ricci soliton satisfying $\mathrm{Ric}+\nabla^2 f=\lambda g$, $\lambda\neq 0$. We prove that constant scalar curvature forces the soliton to be rigid. More precisely, the universal cover splits holomorphically and isometrically as $N^{2k}\times\mathbb{C}^{m-k}$, where $N^{2k}$ is Kähler-Einstein with $\mathrm{Ric}_{g_N}=\lambda g_N$. In the shrinking case the quotient is trivial; in the normalization $\lambda=1/2$ one has $R\equiv k$. The Kähler result rests on a Riemannian rigidity criterion. On a complete nonsteady gradient Ricci soliton, if $\mathcal{L}_{\nabla f}\mathrm{Ric}$ is nonnegative or nonpositive everywhere, then it vanishes and the soliton is rigid; no assumption on the scalar curvature is needed. Along the Ricci flow generated by the soliton, this means that a Ricci tensor that is monotone in time is constant in time, and that this forces rigidity. We also give a direct proof that the pinching $0\leq\mathrm{Ric}\leq\lambda g$ forces constant scalar curvature and radial flatness, yielding the rigidity conclusion through the Petersen-Wylie characterization. The proof combines a weighted cutoff argument with a partial Codazzi symmetry for the Ricci endomorphism.

Industrial environments are increasingly characterized by the tight interaction among physical processes, communication infrastructures, and intelligent applications. In this context, Digital Twins (DTs) have emerged as a key technology for system analysis and optimization. However, existing DT solutions typically focus either on industrial processes or communication networks, while lacking an integrated and application-aware perspective. To fill this gap, this paper proposes a modular DT framework for industrial environments that jointly models physical processes, wireless communications, and application logic within a unified architecture. The feasibility of the proposed framework is experimentally validated through a real-world Proof-of-Concept (PoC) implemented in the BI-REX pilot line, involving a 5G-connected Autonomous Mobile Robot (AMR) transporting hazardous liquids and remotely controlled by an AI-driven application. The proposed DT is used to reproduce the behaviour of the real deployment and to investigate the impact of different placements of the AI application, including on-premise, edge, and remote cloud execution scenarios. Experimental results demonstrate a close agreement between DT predictions and PoC measurements in terms of both network-level metrics, such as Reference Signal Received Power (RSRP) and latency, and end- to-end application metrics, including application-level Round- Trip-Time (RTT). Moreover, the analysis shows how inaccuracies of network modeling can critically affect the feasibility of latency-sensitive industrial control loops, highlighting the potential of integrated DTs as tools for the pre-deployment design and validation of next-generation industrial systems.

The hyperspherical-harmonics (HH) expansion does not require uniform resolution: different components of the wave function converge at very different hyperangular and hyperradial scales. We use active learning to exploit this structure, allowing the calculation to distribute resolution instead of prescribing it. The HH space is decomposed into physically identifiable classes whose hyperangular and hyperradial cutoffs evolve independently; a Gaussian-process surrogate learns the marginal variational gain of each admissible extension, and a cost-aware acquisition selects the next one. The surrogate never replaces the many-body solver: every accepted extension is followed by an explicit solution in the enlarged space, so each reported energy is variational in an explicitly constructed basis. We test the construction on $^3$H and $^3$He with the Argonne $v_{18}$ two-nucleon interaction, without and with the Urbana IX three-nucleon interaction, against uniform HH ladders extended to $K=60$ that reproduce established benchmarks within $1$~keV. Asked for $5$~keV, the adaptive runs reproduce the energy of the uniform $K=60$ space within $4$~keV in every case and certify it, leaving more than $60$% of that space unbuilt. The reduction grows as the requested accuracy is relaxed, to about $88$% at $20$~keV and $95$% at $100$~keV. The selected spaces reproduce the known class hierarchy of the trinucleon, and the one-body radii and magnetic moments are converged at about the $2\times10^{-3}$ level (relative error) in the energy-selected space. A run started from the resolution reached by the mirror nucleus, or by the same nucleus with the two-nucleon interaction alone, certifies a smaller space with a third of the exact solves.

As more users ask AI systems for information, AI-search platforms are becoming a common gateway to web information. Unlike traditional search, which maps keywords to ranked pages, AI search retrieves pages, filters sources, selects citations, and generates answers before users see sources. This selection layer may amplify source bias and turn source choice into a security question. If a platform repeatedly cites domains where new users can publish posts easily, ordinary publication on those domains can become an indirect path into AI-search citations and answer text. Measuring this path is hard: platforms reveal little about citation selection, citations change over time, and the web contains so much background content that later answer changes are hard to attribute to our posts. We present a measurement framework for identifying and measuring this low-barrier publication path, combining cross-platform citation mapping, publication-barrier testing, and marker-controlled publication experiments. Across 10 AI-search platforms, we analyze 17,211 citation instances over 6,356 unique source domains and find: (1) citations concentrate in platform-specific sources, with top-20 domains capturing 20.5–70.8% of per-platform citations, and 15 of 22 tested publication platforms tied to cited source domains had low or medium barriers for both account setup and posting; (2) in our experiments, ordinary publication on preferred platforms changed what entered AI-search outputs: 8 of 10 platforms cited a fabricated concept within seven days, and one high-preference-platform article had greater citation impact than over 20 matched low-preference posts; and (3) this path is commercially available: a $14 GEO purchase produced 13 public posts, and one AI-search platform cited GEO-posted content with our designed markers within one hour.

Structure-preserving numerical methods for gradient flows have been extensively developed when the governing equations and free energies are known. However, accurately predicting dynamics from observational data while preserving intrinsic physical properties remains challenging. Although neural operators, such as Fourier neural operators (FNOs), provide powerful data-driven approximations, their predictions do not inherently guarantee physical structure preservation. To address this challenge, we integrate deep learning with structure-preserving exponential time differencing (ETD) methods. Instead of directly learning the evolution operator, we introduce a *trajectory-inferred free energy* (TIFE), which reconstructs the unknown energy density from observed trajectories through a Duhamel-based learning framework. The learned energy determines the variational force and stabilization parameter, enabling structure-preserving numerical evolution. We further extend the framework to jointly identify the free energy and diffusion coefficient. Under suitable assumptions, we establish the maximum-bound principle, unconditional dissipation of the original learned energy, and error estimates incorporating discretization and inference errors. Numerical experiments demonstrate accurate energy recovery and reliable long-time predictions.

Generative artificial intelligence is reshaping how information is produced, accessed, and circulated, while enabling disinformation campaigns of increasing scale and sophistication. There is currently no clear evidence that fully autonomous AI swarms conduct influence operations at scale, but their enabling capabilities are advancing. We define malicious AI swarms as coordinated, persistent, and adaptive multi-agent systems designed for influence operations, distinguishing them from AI-assisted content production and centrally managed synthetic personas. We examine their implications for hybrid regimes and conflict-affected states in Africa, where institutional constraints and fragile media environments may heighten vulnerability. Drawing on Mali and Ethiopia, we consider how automated influence could infiltrate communities, fabricate consensus, and erode trust in governance and development. The cases illustrate different configurations of state and non-state influence: competing actors in Mali's fragmented information environment, and more organized state-led strategies of narrative management in Ethiopia. African-language and training-data asymmetries may constrain influence capabilities while weakening defensive responses. Hybrid human-AI operations could combine automated scale and adaptation with local knowledge and credibility. This forward-looking risk analysis develops a scenario of increasingly accessible AI-driven coordination, rather than claiming that autonomous swarms are already operating at scale in Africa. We propose a layered governance approach linking technical safeguards to platform accountability, civic institutions, and regional coordination to protect democratic participation, peacebuilding, and development.

Structural measures such as the clustering coefficient and the average shortest-path length characterise how a network is organised. These measures are typically formulated for real, non-negative edge weights. A class of quantum and photonic architectures has complex edge weights instead, whose phases determine whether alternative routes interfere constructively or destructively. These architectures are also bipartite, so triangles are absent and the smallest closed cycle is a four-node square. Existing measures address complex weights and bipartite structure separately: bipartite clustering coefficients quantify clustering through four-node squares but do not contain phase information, while interferometric coefficients retain phase but are defined on triangles. In this work, we define a clustering coefficient for complex-weighted bipartite networks which reduces to the classical bipartite coefficient when the phases vanish, becomes negative when alternative routes cancel, and can be obtained for all nodes from a single sparse matrix product. We show that the phase accumulated around a square is the smallest gauge-invariant carrier of structural phase information in a bipartite network. The clustering coefficient factorises into a topological contribution and a phase contribution, making a bipartite Watts–Strogatz ensemble analytically tractable. For phases uniformly distributed on $[-\Delta,\Delta]$, the mean phase contribution is $(\sin\Delta/\Delta)^4$, independent of node, degree, and topology. We also derive closed-form expressions for the clustering coefficient, open-path visibility, and phase variance. We numerically verify these predictions and further show that phase disorder increases coherent distance.

For a non-empty bounded set $E\subset\R^d$, the upper and lower box dimensions of its orthogonal projections are constant onto almost all $m$-dimensional subspaces. These almost-sure values are called the upper and lower $m$-box dimension profiles of $E$, respectively. We study their relationship with intermediate dimensions, which interpolate between Hausdorff and box dimensions by restricting the relative sizes of covering sets. We establish several equivalent characterizations of both quantities and obtain upper and lower bounds for box dimension profiles in terms of intermediate dimensions, with the lower bounds also involving the upper Assouad spectrum. We further derive explicit formulas for the box dimension profiles in terms of the box dimension and the initial growth rate of the intermediate dimensions, under a suitable condition on the quasi-Assouad dimension.

We consider the pressureless Euler equations in one spatial dimension. These equations model the dynamics of particles on a line that interact only through perfectly inelastic collisions. The flow was first shown to be unique for given initial conditions by Huang and Wang, assuming the entropy and the strong initial continuity of energy conditions hold. In this paper, we prove an analogous uniqueness theorem using a Lagrangian interpretation. The key ingredients are showing that each solution satisfying the entropy condition and an initial kinetic energy bound has this particular type of Lagrangian interpretation and a monotonicity inequality involving conditional expectation.

We introduce a generator-matrix-based geometric approach to the covering radius of $\mathbb F_{q^m}$-linear rank-metric codes. Starting from the $q$-system associated with the code, we define the notion of covering lift and show that the covering radius can be recovered from the hyperplane weights of such lifts. This yields an exact criterion, in terms of the length and effective length of the code, for the covering radius to attain its largest possible value, together with the upper bound $\rho(\mathcal{C})\le \min\{m,n\}-1$ in all remaining cases. We then specialize our approach to $1$-dimensional codes, for which the effective length coincides with the minimum rank distance. In this setting, codes attaining $\rho(\mathcal{C})= \min\{m,n\}-1$ are related to covering lifts that are clubs. More generally, we characterize this case through linear maps into suitable quotient spaces, obtaining equivalent interpretations in terms of generalised evasive subspaces and auxiliary matrix rank-metric codes. These results determine the covering radius for several boundary values of the minimum distance and for all $1$-dimensional codes with extension degree $m\in\{3,4,5\}$. For $m=6$ and every $q$ we provide constructions and, for $q\in\{2,3\}$, computational results showing that $1$-dimensional rank-metric codes with the same length and minimum distance can have different covering radii.

In this paper we consider the $n-3$ dimensional linear section $X_\Pi$ of the Grassmannian $\mathbb G(1,n)$ of lines in $\mathbb P^n$ with a general linear space $\Pi$ of codimension $n+1$, with $n=2k+1$ odd and $n\geq 5$, that is a Calabi-Yau variety. Related to $X_\Pi$ there is a degree $k+1$ Pfaffian hypersurface $\mathcal C_\Pi$ that is the intersection of the dual of $\mathbb G(1,n)$ with $\Pi^\perp$. We prove that $\mathcal C_\Pi$ is rational and we provide a birational parametrization of it. Moreover we study in some detail the congruence of lines of $\mathbb P^n$ given by the centers of the linear complexes parametrised by the points of $\mathcal C_\Pi$.

Media-bridged time series forecasting is expanding to encompass traditional "multivariate" and emerging "multimodal" (e.g., through textual assistance). Existing Time Series Forecasting (TSF) models still rely on paradigm-specific relation, fusion, and temporal modules, hindering a common forecasting backbone across numerical and pre-aligned narrative-flow settings. To explore this, we propose the Multimedia Identity-Aware Prism Network (MIDAPN), a unified spatiotemporal forecasting backbone based on media-general graph adaptation and automatic temporal learning: (1) Following media pre-alignment, our Multimedia Identity-Aware Graph (MIDAG) revisits identity through static essence, dynamic behavior, and latent commonality, inducing affinities that extend variable-specific dependencies across media. Contextual Identity Modulation (CIM) further refines discriminative aggregation. (2) We develop Spectral Prism Convolution (SPConv) to automatically perform hierarchical temporal analysis, balancing coarse trends and fine-grained details. Meanwhile, its Adaptive Search Guidance configures a scale-efficient architecture for temporal-dimension reconstruction. These decoupled yet synergistic components jointly address media identity disentanglement and temporal-scale mismatch. Comprehensive evaluations involving 16 SOTA TSF models across 13 "multivariate" and 12 "multimodal" datasets, alongside targeted long-context comparisons against 14 time series foundation models and fused pretrained language models, demonstrate MIDAPN's consistent superiority and broad shared backbone compatibility. The code is available at \href{https://github.com/leijieruilq/MIDAPN/tree/main}{https://github.com/MIDAPN}.

The brain maintains stable cognition despite continuously changing neural activity. How to extract stable cognitive states from variable neural observations remains a central problem in neural decoding. Existing neural decoding methods map neural observations to predefined external labels based on the stimulus-response principle, often capturing recording-specific spurious correlations. Inspired by how the brain infers the world, and specifically by Bayesian brain theory, we recast neural decoding as cognitive inference constrained by brain-intrinsic priors, yielding high-level meta-neural semantic representations. In decoding experiments spanning five neural recording modalities and three cognitive domains (motor, perception and internal mentation), our cognitive inference method reorganized the geometry of neural observation representations, yielding meta-neural semantic representations that exhibited consistent geometric relationships across cognitive tasks and enabled the recovery of stable cognitive states from variable neural observations. Our work provides an account of how the brain maintains relatively stable cognition despite continual changes in the external environment. Cognitive stability is sustained through cognitive inference from changing neural activity, without requiring fixed neural activity patterns.

Project Greenhouse represents our exploration of a simple thesis: We believe that it is possible to build fully open and sovereign models for agentic search with only modest computational resources. As a first milestone, we describe how to build a competitive pointwise decoder-only reranker using a simple two-step recipe comprising pre-training from scratch followed by supervised fine-tuning, starting only from commonly available datasets. Contrary to the dominant approach in the literature, we do not rely on existing open-weight backbones from third parties, and thus we are fully in control of model training, from end to end. We were able to accomplish the bulk of our experiments using no more than a handful of GPUs. This report articulates the importance and benefits of our approach, and we share artifacts that enable transparent, independent reproduction of all aspects of model training. Beyond data, code, and configurations that capture our efforts, we also release checkpoints for our family of Gaggle models, demonstrating the feasibility of our approach and providing a first step toward validating our broader thesis.

Multipartite entanglement measures probe how quantum correlations are organized beyond what can be inferred from pairwise entropies. We study how this structure is reshaped by a physical boundary in a holographic BCFT. The end-of-the-world brane opens new relative-homology channels for multipartite entanglement, allowing connected bulk webs to reorganize and factorize at scales that need not coincide with any ordinary Ryu–Takayanagi transition. In static AdS$_3$/BCFT$_2$, this produces a characteristic hierarchy of multipartite phases and a nontrivial dependence of genuine multi-entropy on the boundary. In the thermofield-double state, genuinely connected webs can instead be supported through the Einstein–Rosen bridge, leading to growth, reconnection and saturation scales distinct from the usual Hartman–Maldacena transitions. These results show that multipartite entanglement is sensitive not only to extremal-surface areas but also to how bulk entanglement structures are allowed to join, split and terminate in the presence of a boundary.

For persistent and personalized conversational agents, memory systems can enable them to remember, update, and reason over long histories by storing past interactions and retrieving relevant information. Existing memory systems typically follow two paradigms: flat-structured memory and graph-based memory. The former is lightweight but leaves event relations and state updates implicit, while the latter explicitly models memory structure but incurs additional construction cost and introduces irrelevant relations over long histories. To address these limitations, we propose QGMem, a novel memory construction and activation framework motivated by human memory, in which experience is organized into events and query-relevant events are modeled by graph as working memory. QGMem converts long dialogue histories into event-indexed atomic memory units that preserve individual experiences and consolidates related units into dynamic memory traces that retain state trajectories and current states. When a query arrives, hybrid memory retrieval gathers complementary candidate memories, and query-aware reranking activates the most relevant units as a compact working memory. To expose relational dependencies in the working memory and support conflict-aware reasoning, QGMem organizes the working memory as a local graph, which is then encoded as a graph token and provided to the LLM together with the textual working memory to improve evidence utilization during answer generation. Experiments across six benchmarks validate the framework and show consistent gains in retrieval, multi-hop evidence composition, conflict resolution, and ultra-long dialogue reasoning with compact contexts and moderate inference cost.

In this paper we consider random walks moving in a dynamic random environment that is completely resampled with probability $p$ after each step of the random walker. This resampling yields a straightforward regeneration structure and limit theorems, such as a law of large numbers with a limiting speed $v(p)$ that depends on $p$. We investigate several properties of $v(p)$ as a function of $p$, such as differentiability and monotonicity. Similarily, we study the limiting diffusivity $\sigma^2(p)$ in a central limit theorem.

Recent multimodal large language models (MLLMs) increasingly incorporate explainable reasoning for emotion understanding. However, reasoning based mainly on observable affective cues can reduce emotion understanding to superficial cue-label associations, giving rise to the Clever Hans effect. Such shortcuts become unreliable when affective cues are implicit, conflicting across modalities, linguistically misleading, or obscured by redundant details. In contrast, human emotions are shaped by how individuals interpret and evaluate surrounding events beyond observable cues. Inspired by appraisal theories of emotion, we formulate multimodal emotion understanding as a progression from perception to cognitive appraisal, and introduce a dataset, a model, and a benchmark to support this novel paradigm. CogEmo-40K is a large-scale instruction-tuning dataset constructed through a perception-to-appraisal pipeline to elicit evidence-grounded reasoning across six cognitive appraisal dimensions underlying emotion. CogEmo-MoE is a compact sparse MLLM that introduces interleaved MoE blocks for appraisal-specific adaptation, enabling effective appraisal reasoning at a substantially smaller scale than typical emotion MLLMs. CogEmo-Bench introduces an Appraisal Evidence Quality Score (AEQS) to assess cognitive-affective understanding across six complementary appraisal dimensions, addressing the limitation of conventional emotion metrics that evaluate what emotion is predicted but not why it arises. Extensive experiments show that our paradigm not only leads CogEmo-Bench, but also exhibits strong cross-domain generalization. Our findings suggest that perception-to-appraisal reasoning can move beyond surface-level cue-label associations toward more reliable multimodal emotion understanding and closer cognitive alignment between MLLMs and humans.