Papers

Sterile neutrinos remain a well-motivated dark-matter candidate whose cosmological abundance can be enhanced by resonant active–sterile conversion in the presence of a lepton asymmetry in primordial plasma. In the standard picture, a larger initial asymmetry increases the matter potential and can therefore promote resonant production. However, the same increase also shifts the resonance to later stages of the cosmological evolution when the plasma temperature is lower. We show that this delayed production epoch can overlap with the onset of active neutrino oscillations, which redistribute the flavor asymmetries and may substantially reduce, or in some cases nearly erase, the lepton asymmetry needed for a successful resonant sterile-neutrino production. This interplay provides a natural saturation mechanism for the final sterile-neutrino abundance: beyond a certain range of initial asymmetries, increasing the primordial asymmetry no longer leads to a proportional increase in the produced dark-matter density. We identify this effect as an additional constraint on resonant sterile-neutrino production at large lepton asymmetry and discuss its dependence on the flavor structure of the initial asymmetry and on the background cosmological evolution. Our numerical results reveal two order of magnitude range in sterile-active neutrino mixing parameter $\sin^2(2\theta)$ presently consistent with the resonant mechanism of the dark matter sterile neutrino production. It must be investigated by the next generation X-ray telescopes to fully explore this mechanism and corresponding minimal models suggesting sterile neutrinos as viable dark matter. We also consider pion condensation at large flavor lepton asymmetries, which may lead to a first-order QCD phase transition and associated gravitational-wave production in the early Universe.

Reconstructing environments where parts of the scene change between captured image sets poses a challenge for 3D scene reconstruction. We present ChronoFuseGS, a multi-temporal Gaussian Splatting approach that addresses this issue by taking multiple separately trained Gaussian Splatting models, each representing a distinct timestep and partially overlapping in geographic coverage, and merging them into a single combined model. By allowing Gaussians from one timestep to contribute to the reconstruction at other timesteps, our approach leverages data across all captured timesteps to refine persistent parts of the scene. The model supports incremental extension, allowing new timesteps to be added while preserving the existing merged reconstruction. It encodes, for each Gaussian primitive, at which timesteps it contributes to the reconstruction. To support visual exploration of the reconstructed scene, we present a change-aware visualization approach that highlights the parts of the scene that have changed across a user-defined time selection, while preserving the color of persistent parts. Since the persistence encoding operates at the Gaussian primitive level, changes are visualized at sub-object granularity rather than being limited to object-level changes. We evaluate our approach on a real-world outdoor dataset of a flood management area, captured over 7 months across eight recording days and covering seasonal vegetation changes, snow cover, and flooding events, which we make publicly available. Our results demonstrate that the combined model consistently outperforms individually trained single-timestep models in novel-view synthesis quality, recovers structural details absent in the individual reconstructions, and reliably highlights changes in fine details and sub-parts of objects and natural structures.

Accurate modeling of dependent degradation processes is essential for credible reliability assessment and remaining useful life (RUL) prediction in complex systems. Existing dependent degradation models usually describe dependence using correlation-based methods with symmetric characteristics. However, they ignore inherent causal directionality between degradation paths, which may lead to biased reliability evaluation and RUL predictions when physical causality exists. To address this issue, this paper proposes a causality-driven framework for modeling dependent degradation data. Firstly, univariate degradation models with multi-source uncertainties are established for each performance indicator based on the Wiener process. Then, the stable Peter-Clark algorithm is employed to uncover inherent causal relationships between degradation processes, and an uncertainty-aware neural network is employed to quantify causal effects considering uncertainties. Next, univariate degradation predictions and causal predictions are integrated within a Bayesian framework to construct a causally dependent degradation model, and the corresponding loss function is derived for model training. Finally, system reliability and RUL predictions are derived via Monte Carlo simulation. The proposed methodology is validated on the C-MAPSS dataset. Results show that considering inherent causal directionality between degradation processes helps eliminate physically unrealistic degradation behaviors, yielding more accurate degradation and RUL predictions than independent and correlation-based dependent degradation models.

Building a robotic manipulation system requires connecting perception, planning, and control through carefully designed representations and interfaces. VLM code generation offers a way to automate this construction, but independently generated components may operate on incompatible geometric and task-level information. We present Representation-guided Integration of VLM-generated Executable Task programs (RIVET), a framework for generating complete manipulation systems around a shared object-centric representation. The representation combines per-object 6D poses, which preserve the metric information required for action grounding, with a relation graph that exposes the task-level structure required for planning. Guided by this representation, a VLM generates cooperating perception, rendering, relation-inference, and planning programs, each combining task-specific computation with available packages where useful. The resulting programs are authored once for a manipulation domain and reused on unseen start and goal configurations without code regeneration. We evaluate RIVET on cube stacking, tangram rearrangement, and three-dimensional assembly in simulation and on a physical robot, where we achieve 83% overall success rate in the real world by reusing offline-generated systems. Our results demonstrate that representation-guided program generation can adapt a common manipulation framework to tasks with different geometric, relational, and sequential requirements.

Place a queen in each successive column of an infinite $\mathbb{N}\times\mathbb{N}$ chessboard, always choosing the lowest row such that no two queens may attack one another. We prove that the row $q_n$ occupied by the queen in the $n$th column satisfies $q_n=n\phi+O(1)$ or $q_n=n/\phi+O(1)$, where $\phi=(1+\sqrt5)/2$ is the golden ratio.

We describe, with proof, an infinite class of examples of finite sets of integers $A$ for which $\|\widehat{1_A}\|_1\geq (c_\mathcal{R}-o_{|A| \rightarrow \infty}(1))\sqrt{|A|}$ where $c_\mathcal{R}=0.953\dots$.

We present the formalization of Carleson's theorem in the proof assistant Lean. This paper describes the mathematical content, organization of the project, the blueprint, and the main design decisions behind the formalization. It is the result of a large collaborative effort, written and developed in public.

Ultrashort X-ray pulses, as provided by X-ray Free Electron Lasers (XFELs), offer unique opportunities to probe the formation of laser-induced nanostructures with sizes even below the optical diffraction limit, with sub-micometer and sub-picosecond spatial and temporal resolution. Using ultrafast small-angle X-ray scattering (SAXS), we probe the formation of nanoscatterers and nanocracks in reciprocal space as they evolve into different types of laser-induced periodic surface structures (LIPSS) on silica films. Our focus is on the dynamics and evolution of nanoscatterers and high spatial frequency LIPSS (HSFL), with spatial periods below 100 nm. Our experiments include an in-situ analysis of the pulse-by-pulse evolution of HSFL at a fixed sample location, as well as fs time-resolved delay and fluence scans using an ultrafast pump-probe scheme. We find that nanoscatterers and periodic nanocracks form at delay times of hundreds of picoseconds and subsequently evolve into HSFL. At low laser fluences, inter-pulse incubation effects become important during multi-pulse irradiation. Simultaneously, inter-pulse feedback leads to an ordering and improve the regularity of the HSFL. The experiments are complemented by theoretical analyses, including numerical Finite-Difference Time-Domain (FDTD) simulations. These combined analyses allow us to propose a five-stage model of LIPSS formation on silica films upon irradiation with fs-laser pulses.

Many studies expose units to several binary treatments at once and record several correlated outcomes, yet analysts usually estimate one treatment's average effect on one outcome at a time – discarding how treatments interact and how their effects vary across units. We introduce the factorial multivariate Bayesian causal forest, a Bayesian nonparametric model that decomposes the factorial response surface over the treatment lattice into a prognostic sum-of-trees plus one sum-of-trees per main or interaction effect, each with correlated multivariate leaf parameters, estimated jointly with coherent uncertainty. A single indicator-weighted kernel samples every component – the prognostic term is the special case whose indicator is unity – so the two-treatment model, the single-treatment multivariate causal forest, and a general order-r truncation are one and the same sampler. We give the ANOVA/Möbius identification of each estimand, an efficient Rcpp engine, and an interpretability layer using value-suppressing uncertainty maps. In simulations the average-effect estimators are unbiased with nominal coverage, consistent, and robust under misspecification, failing only under unmeasured confounding, which we flag. We illustrate the method on clinical, agricultural and economic data and on a deeper application to the NHANES survey. Software is provided as an R package.

A packing of $n$ congruent balls in $\mathbb{R}^3$ is a family of interior-disjoint Euclidean balls all having the same radius. The contact number of a packing is the number of touching pairs of balls. In this paper we investigate the problem of determining the maximum contact number, $c(n)$, of a packing of $n$ congruent balls in $\mathbb{R}^3$. We first show that all packings of $n$ congruent balls that have a contact number of $c(n)$ are minimally rigid. Furthermore, we show that $c(n)=3n-6$ for $n=6,7,8,$ and $9$. These two results resolve a conjecture of K. Bezdek and Khan. During the proof of the latter result, we also enumerate the contact structures of all packings of $n$ congruent balls with contact number $c(n)$ for $n=6,7,$ and $8$. Additionally, we provide a lower bound construction which shows $c(n)> 6n-6\sqrt[3]{2}n^\frac{2}{3}$ when $n=16k^3-33k^2+24k-6$ where $k\in \mathbb{N}$. We also look at the restricted problem where each ball is centered on the face-centered cubic lattice $A_3$. In this case let $c_{A}(n)$ denote the maximum contact number. We show that $c_{A}(n)\leq 6n-\frac{6}{\sqrt[6]{2}}n^\frac{2}{3}$ for all $n$, and determine the asymptotics of $c_{A}(n)$ to be $c_{A}(n)=6n-(1+o(1))6\sqrt[3]{2}n^\frac{2}{3}$.

Let $E_{A,B}\colon y^2 +Axy + By = x^3$ be an elliptic curve over $\mathbb{Q}$. The $3$-torsion point $(0,0)$ induces a $3$-isogeny $\phi\colon E_{A,B}\to E_{A,B}'$. Assuming $E_{A,B}$ has good reduction at $3$, we construct an explicit $(m+t)\times m$ matrix $M_{A,B}'$ over $\mathbb{F}_3$, whose kernel encodes the dual isogeny Selmer group $\operatorname{Sel}_{\widehat{\phi}}(E_{A,B}')$ modulo the image of the torsion point $(0,0)$. Here, $m = \omega(B) - 1$, and $t$ encodes the global Selmer ratio $3^{t-2}$. We compute $M_{A,B}'$ in various regimes for billions of elliptic curves $E_{A,B}$. Based on our data, and motivated by prevalence of random linear algebraic models throughout number theory, we conjecture that, for fixed $m$ and $t$, the matrices $M_{A,B}'$ become uniformly distributed, and we formulate a corresponding conjecture for the distribution of $\operatorname{Sel}_{\phi}(E_{A,B})$. Our model predicts that, for fixed $t$, the average size of $\operatorname{Sel}_{\phi}(E_{A,B})$ is $1 + 3^t$.

Morphology–control co-design jointly optimizes an agent's body structure and control policy as an integrated embodied system. However, existing methods typically model morphology design and control with separate networks coupled only indirectly through a shared task objective, limiting explicit high-level coordination. Inspired by natural genes that coordinate biological development, we introduce Morphogene, a compact latent blueprint that bridges an agent's body and brain. Through AdaConcat, Morphogene jointly conditions morphology and control generation at the limb level, allowing its variations to induce coordinated changes in both components. Building on this representation, we propose GeCode, which formulates co-design as exploration in the compact Morphogene space. Each Morphogene anchors a local design region in which nearby body–brain designs are explored, while performance-guided updates move these anchors toward promising regions for more efficient exploration of the broader design space. This process combines local refinement with global exploration while preserving body–brain compatibility. Extensive experiments across diverse 2D and 3D co-design tasks demonstrate that GeCode consistently outperforms existing state-of-the-art methods, achieving substantially faster convergence and higher final performance.

We give a deterministic fixed-parameter algorithm for integer cone optimization with convex costs. Given $\tau$ rational polytopes in $\R^{d+k}$, we combine generators whose first $d$ coordinates are integral and whose remaining $k$ coordinates may be fractional. Each generator has a convex cost depending only on its integer coordinates. We minimize the total cost of nonnegative integer combinations subject to arbitrary rational linear constraints on the full typewise aggregates and generator counts. Our algorithm runs in time \[ 2^{\tau\log(\tau+1)\,2^{O(d)}}\cdot |I|^{O(1)}, \] where $|I|$ is the total binary input length. It constructs an optimal representation using at most $2^{2d+1}$ distinct generators per type. The number $k$ of continuous coordinates contributes only to the polynomial input factor. For a single generator type, the running time becomes $2^{2^{O(d)}}\cdot |I|^{O(1)}$. Applications include constructive FPT algorithms for Cone and Polytope Intersection, high-multiplicity bin packing, vector bin packing, and makespan minimization on uniform and unrelated machines. The double-exponential dependence on $d$ is tight for Point-in-Cone and both bin-packing problems assuming $\mathrm{NP}\nsubseteq\mathrm{SUBEXP}$. The mixed-integer formulation also yields additive EPTASs for makespan minimization, machine covering, and envy minimization on identical machines, and additive schemes fixed-parameter tractable in $1/\varepsilon+\tau$ on uniform machines.

We study hyperedge partial duals of finite hypermaps in a purely combinatorial framework, without assuming orientability. A hypermap is represented by three fixed-point-free involutions $(\tau_0,\tau_1,\tau_2)$ on its flag set. We first give an explicit construction of the medial map from this model: $02$-orbits become the medial vertex discs, while $\tau_1$-transpositions become the medial bands; a local orientation system and its twist data then provide a signed rotation description of the medial map. We next prove that the state circles associated with a chosen set of hyperedges are in natural bijection with the vertex orbits of the corresponding partial dual, yielding a crossing-total characterization of all Eulerian hyperedge partial duals. For bipartiteness, the twist data lead to a modified medial map in which inserted bars record the obstruction to a global orientation. We prove that a partial dual is bipartite if and only if its dualized hyperedge set is exactly the set of $c$-type hyperedges identified by an all-crossing orientation of this modified medial map. When the hypermap is orientable, these constructions specialize to the known orientable-hypermap results; when every hyperedge has valence two, they specialize to the ribbon-graph results.

Ordinal acne severity grading requires distinguishing visually similar neighboring grades while jointly weighing holistic facial appearance and localized lesion burden - evidence that most existing approaches collapse into a single opaque representation. We introduce CG-HAF, a global-local fusion framework that instead keeps this evidence explicit: averaged holistic severity probabilities from independently trained classifiers are combined with structured lesion-burden descriptors from an object detector (lesion count, detection confidence, lesion area) into a compact representation, from which a lightweight, interpretable classifier produces the final grade. On a widely used benchmark, this fusion yields a clear, statistically supported improvement over global-evidence-only baselines, with the largest gains on the most severe cases. Testing on an independent dataset with a different grading standard shows that strong within-dataset performance does not transfer automatically, and a follow-up diagnostic attributes much of this gap to mismatched grading criteria rather than detection failure alone. These findings support interpretable global-local fusion as an effective strategy for ordinal acne grading while highlighting criterion alignment as key to cross-dataset portability, with a further illustration of how the resulting severity signal can support transparent, non-diagnostic decision-making in skincare applications.

Continual spatio-temporal forecasting supports traffic management and environmental monitoring under evolving dynamics and expanding sensor networks. However, conventional graph-based continual learning methods tie forecasting representations to the current sensor layout, so sensor expansion can alter the representation of learned spatial relationships. Our key insight is that sensor expansion changes the evidence available about a process without necessarily changing the dynamics to be learned. We propose STFO (Spatio-Temporal Field Operator), which parameterizes forecasting knowledge as a shared field-evolution operator and handles changing sensor layouts through observation and query interfaces. Normalized coordinate-based aggregation lifts irregular sensor histories onto a fixed latent grid, enabling reuse of learned spatial maps across observation sets without sensor-specific parameters. To accommodate process drift, a spectral descriptor summarizes variation across spatial scales and conditions Fourier propagation and attention to adapt operator responses to the current spatial regime. Coordinate-based decoding queries the evolved field at sensor locations and combines spatial corrections with local-history predictions. Experiments on PEMS-Stream, CA-Stream, and AIR-Stream demonstrate state-of-the-art average forecasting performance. STFO-Large reduces average MAE over DOL by 8.4% on PEMS-Stream and 4.7% on CA-Stream. Our code is available at https://github.com/Xielewei/Spatio-Temporal-Field-Operator.

We characterize long-term hardness behavior in 29 ultraluminous X-ray sources (ULXs) using over 3500 archival Swift/XRT observations. We classify observations as soft-band dominated, hard-band dominated, or uncertain from the hardness ratio (HR; 1.5–10/0.3–1.5 keV count rates) and its confidence bounds. We derive $f_{\rm soft}$ and $f_{\rm hard}$ and define the Crossing Index (CI) as the fraction of consecutive confidently classified observations on opposite sides of $HR=1$. We analyze these descriptors with rule-based classification and unsupervised hierarchical clustering. The rule-based analysis identifies eight soft-band-dominated, five hard-band-dominated, and 16 mixed-band systems; clustering yields a similar three-group structure. Known or candidate accretor types tend to occupy band-dominated groups, whereas unknown accretors are more common in the mixed-band group. All four sources with reported long-term X-ray modulation discussed in an orbital or super-orbital context are classified as mixed-band. The three-cluster solution is preferred in 25 of 29 leave-one-out trials, while bootstrap tests find the band-dominated groups more stable than the mixed-band group. As the classification uses observed count rates, the groups represent Swift/XRT signatures rather than intrinsic spectral states. Response-folding tests show that modeled HR depends on absorption column density and photon index. Bayesian reanalysis of a representative low-count source reduces CI from 56.3% to 38.5% while preserving its mixed-band classification. Random downsampling produces no systematic shift in CI; halving the observations changes median CI by at most $\sim10\%$ across four sources. Larger samples with independent absorption constraints can test links among these observational classes, accretor type, long-term periodic modulation, and other X-ray properties.

In this work we study if a robotic hand using proprioception alone can grasp diverse objects with no visual observation. We present a modular dexterous grasping architecture that separates global arm motion from local contact control. An independently controlled arm guides the hand toward the object, while a reinforcement learning policy grasps and stabilizes it using only hand proprioceptive feedback. We call this a blind grasp reflex: grasping without images, object poses, or geometric observations. A learned stable-grasp score determines when the object is securely held, allowing the arm to begin post-grasp manipulation. This separation makes grasping a reusable hand-level skill that can be combined with independently designed arm controllers for various manipulation tasks. Experiments in simulation and on hardware demonstrate robust blind grasping across diverse objects and seamless composition with a range of arm controllers. Moreover, despite never observing contact geometry, the learned grasp score closely aligns with an independent physics-based measure of grasp stability. The resulting approach follows a simple principle: see to reach, feel to grasp. Project page: https://blindgraspreflex.github.io.

An algebraic $q$-difference equation for $|q|>1$ is considered. A sufficient condition is proposed for the convergence of a solution to such an equation in the form of a Dulac series within a sector centered at the origin. An example illustrating the application of this sufficient condition is provided.

We present a detailed feasibility study of exotic muon decay searches using the muon beam at the High Intensity heavy-ion Accelerator Facility (HIAF). For the decay channel mu to eX0, where X0 denotes a new boson featuring LFV couplings, the produced electron carries very low energy when the mass of X0 approaches the muon mass, making detection challenging. Nevertheless, the high-energy beam at HIAF can boost these electrons up to a measurable energy range. The detector system proposed here comprises front-end scintillators for particle identification (PID) and a main detector consisting of an RPC stack to track decay products and a scintillator for energy measurement. Leveraging the kinematic properties of low momenta and small emission angles of signal electrons, we employ optimized momentum and angular acceptance cuts on the RPC detector to efficiently suppress background events. With this strategy, we obtain an upper limit of 1e-5 on the branching ratio of the exotic muon decay at 95% confidence level, achieving state-of-the-art sensitivity.

The present work proposes a new operator splitting technique. The method leverages ideas used in PDE-constrained domain decomposition, adapting them to the objective of decoupling equations defined on the same geometric domain. The methodology is applied here to the resolution of coupled flow and deformation in the linear quasi-static Biot problem. An overview of its main characteristics is provided, also in comparison with the well established fixed-stress split for poroelasticity.

In a recent paper Nahum argued on the basis of a renormalization-group analysis that, contrary to standard lore, ground states of 1D spin chains with short-range interactions can spontaneously break U(1) “easy-plane” spin rotation symmetry. True long-range order of $(S^x,S^y)$ arises at the phase transition between two quasi-long-range-ordered phases. Here we present detailed numerical results for an anisotropic spin-1 chain model. We find a magnetization exponent $\beta\approx0.29$, consistent with the $\epsilon$-expansion prediction $\beta\approx0.28$, a divergence of the Luttinger parameter on approaching the transition from both sides with an exponent that we relate to Nahum's transition, and finite-size spectra consistent with a dynamical critical exponent $z\approx2$. Our results support the critical behavior predicted by Nahum's field theory.

AI agents combine language models with external data and tools that can modify files, call APIs, or execute code. Security failures can arise when adversarial content changes an agent's tool use or when the surrounding software contains vulnerabilities such as path traversal or command injection. We study authorized white-box pre-deployment auditing, where the auditor has access to the target repository and a controlled runtime, but successful attacks must still act through the task-defined attacker interface and be confirmed by an external verifier. We present AgentXploit, a two-role auditing system that separates repository-level attack-path discovery from runtime exploitation. The Analyzer Agent traces attacker-controlled inputs to sensitive operations and records code-supported candidate attack paths; the Exploiter Agent turns these paths into concrete attacks and revises them using runtime feedback. We also introduce AgentXploit-Bench, containing 72 reproducible vulnerabilities across 12 open-source AI-agent systems and frameworks. Across three runs, AgentXploit reaches 59.3% end-to-end success, compared with 38.4% for Codex. Under a token-budget-matched comparison, Codex reaches 46.3%. On AgentDojo, where injection points are provided, the Exploiter Agent reaches 79.2% attack success versus 52.7% for AgentVigil. These results highlight repository discovery and runtime exploitation as distinct challenges in end-to-end agent security auditing.

We propose that the application of static electric fields to reactive CaNH$_2$ molecules can result in large ratios of elastic to lossy collisions at ultracold temperatures. In the first rotationally excited manifold, the asymmetric top structure of CaNH$_2$ results in a $\approx 226$ MHz splitting of the opposite parity $K$-doublet states. This modest splitting allows electric fields below a kilovolt per centimeter to polarize the molecules and induce long-range dipole-dipole interactions. By increasing the electric field, we find that dipole-dipole attraction can become sufficiently large to induce weakly bound octatomic field-linked states of [CaNH$_2$]$_2$. In the absence of these field-linked states, we identify select molecular states where dipolar interactions allow elastic collisions to dominate over lossy collisions by up to three orders of magnitude, while maintaining loss rate coefficients at the $10^{-14}$ cm$^3$s$^{-1}$ level. These results establish that evaporative cooling is a viable route to quantum degenerate gases of laser coolable asymmetric top molecules. Access to such low entropy ensembles of asymmetric top molecules holds promise to push the frontiers of quantum simulation, state-resolved chemistry, and searches for beyond the Standard Model physics.

Equations of motion are derived for excitation exchange between two atoms having arbitrary angular momentum.

Unobserved common causes are pervasive in real-world time series and can induce spurious associations that causal discovery methods mistake for direct edges. We propose LUCID (Learning Under Confounding for Inference and Discovery, a regime-adaptive deconfounding layer that first estimates the confounding regime from data using a Marčenko–Pastur spectral router, then applies a deconfounding strategy matched to that regime. When the spectrum indicates pervasive factor confounding, LUCID attenuates factor-dominated variation and recovers contemporaneous (lag-$0$) structure from the resulting innovations, with edge selection calibrated against a data-driven edge-free null. Rather than being tied to a particular discovery algorithm, it can wrap existing discovery engines; we demonstrate consistent improvements across three such methods. On a diverse synthetic out-of-distribution benchmark spanning changes in confounder strength and sparsity, loading density, lag structure, volatility dynamics, edge heterogeneity, persistence, intermittency, and tail behavior, LUCID achieves the best family-weighted directed, lag-resolved graph $F_1$ ($0.60$), improving over the strongest baseline by $0.19$ absolute ($\approx\!46\%$ relative). Its advantage widens relative to looser lag-collapsed scoring, and remains robust under intermittent and heavy-tailed confounding. Code reproducing the method, the benchmark generators, and every reported experiment is available at https://github.com/bloomberg/causal-ts.

Cytopathology detection requires open-vocabulary recognition because cellular categories are fine-grained, long-tailed, and continuously evolving across different organ systems. However, existing cytology detectors are mostly single-domain and closed-set, and there is still no unified benchmark for evaluating open-vocabulary cytopathology detection. We present PentaCyto, a multi-domain benchmark covering cervical, urinary, respiratory, serous fluid, and thyroid cytology, with 24 base categories and 9 held-out novel categories. Each category is associated with structured cytomorphology prompts that describe diagnostic morphological attributes and provide clinically grounded textual knowledge. We further propose CytoSPM, an efficient detector based on a decoupled two-stage design. It first extracts reusable class-agnostic visual representations, and then performs class-aware structural prompt matching with class names and cytomorphology prompts. On PentaCyto, CytoSPM outperforms existing methods in novel-category detection and open-vocabulary detection while maintaining efficient inference.

Vision-Language-Action models (VLAs), while showing strong potential for robot control, require massive amounts of high-quality imitation learning data. Moreover, the absence of an explicit world model casts further doubt on their control capabilities. In this concept paper, we propose a novel architecture that addresses sample efficiency in VLAs by training a predictive world model on the embedding space of the VLA's vision encoder. We hypothesize that these embeddings are action-relevant and usable for future prediction. To this end, we propose using the suggested architecture to investigate how well these embeddings predict the future based on actions, as the inability to do so would mark a key limitation of VLA architectures: the lack of a non-lossy implicit world model to simulate real-world dynamics. The proposed architecture differs from the standard world model dynamics as the loss comes from the embedding space rather than the pixel space, similar to joint embedding predictive architectures. Furthermore, the trained world model can be utilized for short-term planning tasks by sampling VLA actions given goal images. We intend to examine the richness of vision embeddings in VLAs and reduce their high data requirements through a world model that can also generate plans during inference.

We construct a causal query compiler for time-dependent Lindblad dynamics and establish query and gate bounds under distinct input models. With coherent block encodings of the Hamiltonian and individual jump operators, a Lipschitz generator can be simulated to diamond-norm error $\varepsilon$ using $O(1+\tau+\log(1/\varepsilon)/\log(e+\log(1/\varepsilon)/\tau))$ controlled queries, where $\tau=T(\alpha_H+\sum_\mu\alpha_\mu^2)$. This bound is independent of the time-grid size and is worst-case optimal for $\tau\ge1$. Weak active entry and clean return yield a factorial history tail; combining two reuse lengths removes the auxiliary input. A three-call polar correction supplies the Lindblad cells. The compiler also characterizes the minimum oracle action of bounded adaptive Markovian protocols by quantum query complexity and the general adversary bound, up to constant factors. For finite-range dynamics on $n$ lattice sites, efficient coherent evaluation of the local matrices and fixed local parameters give $O(n(T+1)\operatorname{polylog}X)$ elementary gates, where $X=\max\{e,n(T+1)(1+K_t+B)/\varepsilon\}$, for piecewise Hölder generators with fixed exponent, local variation bound $K_t$, and at most $B$ breakpoints per term. The jumps need not commute. This gate bound includes evaluation, arithmetic, selection, routing, and environment storage. Its proof combines a spatial decomposition on a common bath, compilation on occupied inputs, and routing on compressed records. Reusing bath storage between short segments gives depth $(T+1)\operatorname{polylog}X$ and space $n\operatorname{polylog}X$.

Measurement error in the Watanabe-Sagawa-Ueda formulation is the excess estimation error of a classical record over the quantum statistical limit. We show that, at the quantum-instrument level, this excess has an exact decomposition. The record-only error matrix equals a residual-aware error matrix plus a positive Schur-complement term fixed by Fisher information left in the unread post-measurement system. Thus measurement error separates into intrinsic instrument loss and record-restriction loss