Combinatorics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 9 November
A connected $k$-regular graph $G$ on $n$ vertices is co-edge-regular if any two distinct non-adjacent vertices have the same number of common neighbors. For adjacent vertices $u$ and $v$, let $\lambda(u,v)=|N(u)\cap N(v)|$. Ge and Koolen asked whether the theorem of Neumaier on strongly regular graphs with fixed smallest eigenvalue extends to co-edge-regular graphs with exactly four distinct eigenvalues in which $\lambda(u,v)$ takes exactly two values over the edges $uv$. In this paper, we answer this question. For every integer $t\ge2$ there is $n_t$ such that if $G$ is such a graph with smallest eigenvalue at least $-t$, $n\ge n_t$, and $k<n-2-(t-1)^2/4$, then either $G$ is a clique extension of a strongly regular graph, or $G$ is obtained from a strongly regular graph with a Hoffman coloring by adding all edges inside the color classes. Consequently, $G$ is a clique extension of a strongly regular graph, a $p\times q$ grid with $p\ne q$, or the block graph of a resolvable Steiner $2$-design in which every parallel class is made into a clique.
Computation and Language
Wrong with Confidence: Measuring Hallucination, Overconfidence and Identity-Conditioned Behaviour in Large Language Models
Large language models answer underspecified questions readily, and accuracy-only evaluation rewards guessing over abstaining. Observational studies link query underspecification to hallucination, but they cannot separate the effect of missing information from item difficulty, and they rarely ask what the model tells the user. We report a pre-registered, paired behavioural study of five open-weight models (Gemma 3 27B, Gemma 4 31B, gpt-oss-120b, Qwen3-Next-80B-A3B and Nemotron 3 Super). Each of 142 factual questions was posed in a fully specified form and in a referentially underspecified form with the same answer key, and three probes measured identity-conditioned behaviour. Two raters scored 5,620 responses under a fixed rubric; effects were estimated with mixed-effects models. Without the referent, the models sought clarification on 33.7% of prompts and answered the rest. Among attempted answers, the error rate rose from 39.1% to 52.0% (mixed-effects odds ratio 4.12, 95% CI 2.76-6.14), and specialist-domain questions failed more often than general ones in every model. Unelicited verbal uncertainty did not track this risk: 75.3% of wrong answers carried no hedge, wrong answers to underspecified prompts were hedged no more often than wrong answers to clear prompts (27.3% vs 22.3%), and hedging separated correct from incorrect answers only weakly (AUROC 0.56). Gendered pronouns followed occupational stereotypes whenever a model committed to one; any stated user identity lengthened answers by about 45% relative to an anonymous user; claimed expertise did not reduce safety caveats. The models either ask or answer as if the question were well posed, and the wording of a guess does not reveal that it is one. Evaluations should score clarification as a distinct outcome and check label validity as well as rater agreement.
Read the full abstract
Computer Vision
Latent Fusion: Channel Stacking for Structure-Preserving Image Editing
Instruction-based image editing models must condition on the source image, either by sequence concatenation, which appends source tokens to the target sequence (FLUX.1 Kontext), or by channel-wise concatenation, which stacks the source latent onto the noisy latent and widens the input projection (InstructPix2Pix, FLUX.1 Fill). The FLUX.1 Kontext report found channel-wise concatenation worse in initial experiments at 1K resolution with a 12B model and a large corpus. We explore channel-wise concatenation as the lower-cost option for a common production regime: structure-preserving edits that change appearance (sky, exposure, foliage, cleanup) but not layout, about 1,400 training pairs, a 4B non-distilled backbone, and 2K frames. It keeps the sequence at its original length, where sequence concatenation doubles it and roughly quadruples attention cost, and needs only single-scale classifier-free guidance (two forward passes per step instead of three). Starting from FLUX.2 Klein 4B base, we widen only the input projection and train both schemes with the same trainer, data, and hyperparameters. Channel-wise concatenation keeps the output pixel-aligned with the source; in our runs sequence concatenation, which must learn the correspondence between 10,880 target and 10,880 source tokens from fewer than 1,500 pairs, displaced structure, an observation we report without claiming a general ranking. On 29 held-out images the editor reaches 20.3 dB PSNR and 0.198 LPIPS against the human edit (unedited source: 15.0 dB), and counterfactual conditioning tests show it largely learned a global grade of the source rather than a selective edit, which we report as the open problem of this recipe. We also compare against an SD3.5-Medium editor trained with the same surgery and report the full training recipe, including failure modes. Code: https://github.com/complete-dope/LatentFusion
Read the full abstract
General Mathematics
Zariski Topology of (Krasner) Hypermodules
Let $R$ be a commutative Krasner hyperring with identity and $M$ an $R$-hypermodule. We consider the prime spectrum $\operatorname{Spec}_R(M)$, consisting of all prime subhypermodules of $M$. In this paper, we introduce a Zariski topology on $\operatorname{Spec}_R(M)$ and investigate its fundamental topological properties, including irreducibility, connectedness, compactness, and separation axioms, providing necessary and sufficient conditions for each. In particular, we prove that $\operatorname{Spec}_R(M)$ is irreducible if and only if the nilradical of $M$ is a prime hyperideal, and that it is connected precisely when the underlying hyperring is not a nontrivial direct product. We further investigate the functoriality of the Zariski topology in the category of Krasner hyperrings and hypermodules. Moreover, using the fundamental relation $\epsilon^{*}$, the smallest regular relation on $M$ for which the quotient $\overline{M}=M/\epsilon^{*}$ becomes an $\overline{R}=R/\Gamma^{*}$-module, we establish a correspondence between the Zariski topologies of $M$ and $\overline{M}$. We show that this correspondence is functorial and thereby relate the hypermodule-theoretic setting to the classical theory of modules over rings.
Read the full abstract
Computer Vision
Detection Is Not the Finish Line: Multiscale Decoding and Constraint-Aware Global Association for Multi-Camera 3D Tracking
A query-based multi-view 3D detector can transfer well to a new calibrated camera network and still produce badly fragmented tracks: on the AI City Challenge 2026 Track 1 validation set, the frozen Sparse4D detector behind the official challenge winner splits each Person identity into a median of eight fragments, and half of the gaps between fragments exceed the horizon of a causal tracker. We show that most of the remaining headroom of such a system lies after detection. Keeping the public detector frozen, we add (i) class-wise source selection between two input scales, which recovers small and distant people from a 0.75x pass while leaving large objects at the scale the detector was trained for; (ii) a geometry-to-appearance bridge that projects every 3D box into all calibrated views, matches it to 2D person detections, and aggregates robust track-level re-identification descriptors; and (iii) constraint-aware consolidation, which combines a causal short-gap stitch with offline centroid-linkage clustering under two hard constraints, temporal exclusion and metric 3D reachability, plus class-aware long-gap stitching for rigid objects. Every association stage changes object identities only. On the challenge evaluation server the system scores 62.3947 3D HOTA (DetA 57.4501, AssA 56.8131, LocA 80.1272), first on the general leaderboard snapshot of 30 October 2026 (second at the time of writing, after a competitor's detector update) and 5.63 points above our reproduction of the official winner; the constrained association stage alone contributes +3.73 HOTA through a +5.49 AssA gain at unchanged DetA. Controlled fixed-box validation studies, a constraint ablation, sensitivity curves, and a catalogue of negative results support the design.
Read the full abstract
Data Analysis and Statistics
Some Considerations on Measuring and Comparing Things
Measurements and relationships between measurements are basic aspects underlying science and technology. In the present text, we present and discuss some of the approaches for measuring, normalizing, and comparing entities represented in terms of a set of measurements. Three main categories of measured values (constant, direct, or inverse) are proposed, underlaid by the relationship between data variation or intervals of interest and magnitude of the quantifications. For example, in certain cases the variation of measurements can be expected to be proportional to the relative magnitude (e.g.~when the magnitudes extend over long ranges). The considered approaches to comparing measurements include the Euclidean distance, three types of proximity based on the latter, as well as the Jaccard and coincidence indices. The presentation is complemented by graphical illustrations.
Read the full abstract
Computer Vision
Acting by Intent, Reacting by Sight: Visual Conditioned Reflexes for Real-Time VLM Agents
Vision-Language Models (VLMs) have shown strong capabilities as generalist agents that perceive, reason, and act in interactive environments. Yet real-time environments such as video games demand both deliberate control for exploration and split-second reactions to fast-moving hazards. Existing agents struggle to handle both: relying solely on direct VLM actions is too slow for sudden events, whereas delegating execution to fixed scripts leaves the agent blind at the moment of action, leading to frequent misfires whenever the scene shifts. Humans overcome this by pairing normal conscious control with conditioned reflexes-while the mind handles deliberate actions and shapes reflex rules, the reflex itself still relies on live vision to make the final instant decision. Emulating this mechanism, we propose ViCoR (Visual Conditioned Reflex). Alongside normal gameplay actions, the main VLM defines and updates lightweight situational guidance for recurring events, while a fast visual reflex branch (invoking the VLM in no-thinking mode) inspects live local views with motion cues to make the final sub-second action decisions. By offloading 67.8% of high-frequency reactions to the reflex branch (680 ms per check), ViCoR reduces average per-action decision latency by 2.6x (from 14.0 s to 5.4 s) and cuts token consumption by 36.9%. Across 17 diverse 2D and 3D games in the GameWorld benchmark, ViCoR achieves state-of-the-art performance, improving task success rate by +11.8% (from 52.9% to 64.7%) and average normalized progress by +13.9% (from 66.5% to 80.4%) over the strongest deliberative VLM and script baselines. In fast-paced real-time games, ViCoR refines its reflexes across episodes to boost progress by 1.67x, and even with a low-effort main planner, surpasses high-effort deliberative baselines by +8.1% at one-third of the token cost.
Read the full abstract
General Mathematics
The Maximum Number of Sets for 16 Cards is 26
In SET, each card has four properties with three values each, and three cards form a set when every property is all the same or all different. We prove that the maximum number of sets among $13$, $14$, $15$, and $16$ cards is $16$, $19$, $23$, and $26$, respectively. Viewing cards as points of $\mathbb{F}_3^4$ and sets as lines, we show that a configuration that leaves one of the $40$ line directions of $\mathbb{F}_3^4$ unused projects into $\mathbb{F}_3^3$ without losing a set, which reduces the problem to three properties, where exhaustive search is feasible. We force an unused direction by counting repeated directions in the planes through a card on many sets, and treat configurations containing a magic square separately.
Read the full abstract
Fluid Dynamics
Terahertz and infrared links in controlled non-Kolmogorov turbulence
Hybrid terahertz/infrared (THz/IR) links are usually modeled with Kolmogorov turbulence, but whether the turbulence spectral type matters equally for both bands has not been investigated. We transmit co-propagating 625 GHz and 1550 nm links through laboratory-controlled turbulence while a Mach-Zehnder interferometer and thermistors independently measure its spectral exponent and structure parameter on the shared path. Inlet-air temperature, not airflow speed, drives the exponent from Kolmogorov-like to non-Kolmogorov. At comparable structure parameters, this change reverses which model better reproduces the IR attenuation at the nominal inner scale, whereas the THz attenuation is insensitive to it. With type-matched models, predicted and measured bit-error-rate trends agree within 0.14 (IR) and 0.06 (THz) decades. IR predictions therefore require the spectral shape and inner scale, not turbulence strength alone.
Read the full abstract
Complex Variables
Calculation of Stiefel fractions via Riemann thetas
Stiefel fraction is a deformation of Zolotarev one which allows to solve certain three band extentions of Zolotarev approximation problem. We propose an efficient computational approach for Stiefel's solution based on genus two Riemann theta functions. This allows us to visualise the domain in the space of three band problems corresponding to Stiefel's solutions.
Read the full abstract
General Mathematics
Counting Second-Order Schreier Sets with Minimum 2
A finite nonempty subset of the positive integers is first-order Schreier if the minimum element of that set is greater than or equal to the size of the set. We can generalize the first-order Schreier sets recursively to higher orders, and we specifically investigate the second-order Schreier sets. We show a combinatorial relationship between the count of second-order Schreier sets with minimum element $2$ and the index shift of a convolution of the Fibonacci numbers (given by Sequence A023610 in the OEIS).
Read the full abstract
Computational Engineering
The norm, not the labels: training and evaluating finite-element surrogates for stress analysis in the energy norm
Finite-element surrogates predict the stresses of a new design in one evaluation, but are usually trained on solutions with a displacement loss and compared by it, although engineering decisions rest on stresses. Application in engineering: stress analysis of perforated plates and solids with cavities, parametric families in which holes concentrate stress. Contribution to artificial intelligence: training and evaluating such surrogates in the energy norm, with or without solutions, and screening their predictions without solutions. For linear elastostatics, minimising the assembled discrete potential energy is exactly supervised regression in this norm: the energy gap of a prediction is the strain energy of its stress error. A displacement loss under-weights the stiff modes that carry the strains. In pre-registered comparisons, changing only the norm of the supervised loss cut a three-dimensional transformer's median energy-norm error 7.5-fold and its median von Mises error 6.1-fold; in two dimensions it more than halved the von Mises error, also on held-out instances. In three dimensions, the transformer trained on the energy alone came within a factor of 1.3 of the energy-norm-supervised one in median energy-norm and von Mises error. It was more accurate than the displacement-supervised one on all five metrics in every seed. Given the stiffness matrix, a Ritz projection was more accurate than the label-free transformer, in the seed means, on every metric but peak stress. No network transferred to a substantially finer mesh. The energy ranks predictions of a load case, and its sign flags those worse than the zero field.
Read the full abstract
Data Structures and Algorithms
Beyond the First Hit: Towards Tight Guarantees for $k$-means++
Let $R_k$ be the worst-case ratio between the expected cost of the $k$ seeds chosen by $k$-means++ and the optimal $k$-means cost. In their seminal paper introducing $k$-means++, Arthur and Vassilvitskii (SODA 2007) showed that $2\ln k\le R_k\le8(\ln k+2)$, and Makarychev, Reddy and Shan (NeurIPS 2020) improved the upper bound to $5(\ln k+2)$. We prove $R_k\le\frac{1163}{233}+2.108\,H_{k-2}$ for every $k\ge3$ by a computer-assisted argument, and $R_k\le5+2.3139\,H_{k-2}$ for every $k\ge2$ by a short analytic argument; here $H_j$ is the $j$th harmonic number. Since $H_{k-2}\le1+\ln k$, this lowers the leading coefficient of $\ln k$ from $5$ to $2.108$. We also determine $R_2=5$ and $R_3=185/32$ exactly; both are suprema that are approached but not attained. On the lower side we prove $R_k\ge2(1+H_{k-1})$ for every $k$, and give a second family of lower bounds that is larger than this for $2\le k\le18$. Previous analyses bound the expected cost of a cluster just after it is first sampled by a constant times its optimal cost, and then charge that cost for the rest of the run. We instead track how a cluster's cost evolves after its first hit: a cluster left with a large residual cost is likely to be sampled again before it accumulates much of the logarithmic charge, and each repeat repairs it.
Read the full abstract
Statistics Applications
Strategic Causal Policy Learning with Manipulable Eligibility
The standard rule in unconstrained causal policy learning is simple: treat individuals with nonnegative CATEs. But this rule can fail when treatment eligibility is strategically manipulable. After a policy is announced, individuals may change the covariates used for assignment without changing their underlying treatment effects. We study how such responses affect treatment allocation and welfare. Using an access-burden representation, we characterize strategic treatment and decompose welfare loss into missed beneficial treatment and harmful treatment. For linear CATEs and quadratic manipulation costs, we derive closed-form threshold corrections. When manipulation costs are homogeneous, a single correction recovers the ideal positive-CATE allocation. With heterogeneous costs, welfare maximization, pointwise safety, and minimax group welfare regret can favor different thresholds. Simulations illustrate these tradeoffs and examine the effects of estimating treatment effects and behavioral responses.
Read the full abstract
Computational Engineering
MechFoundry: An agentic loop that poses and performs engineering analyses and optimization of real-world objects
AI agents can automate software engineering because a compiler or a test suite checks that the code does what the user or the agent intended. In engineering analysis, however, the steps that precede the computation have few easily verifiable checks: modeling a design or an object, deciding how it is used, where supports and loads are applied, and what material each part is made of. A human expert therefore makes these decisions, carries out the steps, and judges the result, one object at a time. We present MechFoundry, an agentic loop coupled with bespoke software that automates structural analyses of real-world objects without a human in the loop. From an object's lexical description, it generates a shape, poses a loading case relevant to how the object is used, assigns a material to each part, solves the nonlinear structural response, and improves the design against the loads it carries. A differentiable GPU solver built on an adaptive octree that need not conform to the surface enables the transition from shape generation to simulation and optimization. Across 50 everyday and engineering objects, the agent posed and solved a loading case in 6.5 minutes and for US$1.3 per object on average. Without a test suite to check against, we had language-model reviewers grade each session, as a first indication of where the agent's engineering judgment holds and where it fails.
Read the full abstract
Statistical Mechanics
Quantitative Gaussian limits of periodic O(N) models at and below the critical temperature
We prove a volume-uniform Gaussian approximation for local spin coordinates in the zero-field nearest-neighbour $O(N)$ model on periodic boxes of side $M$ in $d\ge3$, at and below the critical temperature of the limiting spherical model. On a fixed spatial window, for sufficiently large $M,N$, the unnormalized $1$-Wasserstein error for the first $p\le N$ coordinates is at most $C\sqrt p (N^{-1/2}+s_M)$, where $s_M\to0$ is the spectral shift of the finite-volume mean spherical model. The estimate is uniform on compact ranges of positive directional couplings and inverse temperatures. The target has independent coordinate fields, each a scaled anisotropic Gaussian free field with an independent spatially constant Gaussian term below criticality. On a window containing a nearest-neighbour star (a site and its $d$ positive-direction neighbours), the error tends to zero as $M,N\to\infty$ exactly when $p=o(N)$ and $\sqrt p s_M\to0$. For any choice of periodic infinite-volume accumulation states, the error on any nonempty fixed window tends to zero as $N\to\infty$ exactly when $p=o(N)$. In $d\ge5$, when the inverse temperature exceeds its critical value by at least a fixed positive amount, the bound improves to $C(p/N+\sqrt p M^{-d})$. For fixed $p$, the sharp order is $N^{-1}+M^{-d}$ on star-containing windows and $N^{-1}$ on every nonempty fixed window in periodic accumulation states.
Read the full abstract
Number Theory
Divergent Trajectories in the $3x+1$ Problem and a ${\em S}$- Units Equation
In this note we assume the existence of divergent trajectories (if any) under the Collatz map and investigate the consequences of this assumption, even though the Conjecture of Collatz seems to be true. In a trajectory we show that the Diophantine exponential equation $x + y = w 2^n$ in the integers $x, y, w, n$ plays an important role. In this context the recent strong finiteness theorem of Bennett and Billerey on $S-Units has consequences on the set of prime numbers generated by a collection of local minima, called wave-minima (defined ad hoc for simplicity's sake) in a divergent trajectory. It allows to define canonically a density of a trajectory and some conjectures and open problems related to these prime numbers.
Read the full abstract
Genomics
Deep learning-guided discovery reveals the diversity and mutational potential of environmental beta-lactamases
Beta-lactamases are a major mechanism of bacterial resistance to beta-lactam antibiotics; however, their functional diversity and mutational potential in environmental microbial communities remain partially understood. This study aimed to characterize environmental beta-lactamase diversity and investigate how sequence variation influences antibiotic resistance. By integrating protein language model (PLMs) with ensemble learning, we identified more than 46,000 putative beta-lactamases across all four Ambler classes from environmental metagenomes, widely distributed across diverse habitats, with distinct habitat-associated class compositions. The candidates sharing approximately 30% sequence identity with known beta-lactamases conferred resistance to multiple beta-lactam antibiotics when expressed in Escherichia coli. Combining inverse folding model with PLMs further identified mutations that altered resistance profiles, including substitutions that increased resistance to carbapenems. Predicted high-fitness sites exhibited greater natural sequence variability overall, with higher variability in wastewater-derived than ocean-derived enzymes. These findings reveal undercharacterized beta-lactamase diversity and substantial mutational flexibility in environmental microbiomes, providing a basis for investigating their potential contribution to the diversification of antibiotic resistance.
Read the full abstract
Data Analysis and Statistics
Learning to Calibrate Jet Energies
Hadronic jets are produced copiously at the Large Hadron Collider, and are a vital element of its physics program. Accurate reconstruction of their energies is therefore crucial, including correction for pile-up, detector response and data-simulation differences. Currently, each correction is performed by fitting parametric functional forms, an approach which is effective but can be fragile and challenging to extend. We present a prototype machine-learning pipeline which performs pile-up dependency removal, absolute corrections and data-mc corrections. No functional form is imposed on the correction, additional features can be easily added, and the pipeline matches or exceeds the performance of the functional approach at nearly every stage.
Read the full abstract
Computer Vision
Stylization as Geodesic Navigation on Diffusion Manifolds
Facial stylization asks for a delicate transformation: the image should look unmistakably artistic, yet the person should remain unmistakably the same. Existing training-free and instruction-driven methods often fail to expose such control, either drifting away from identity or producing only weak stylization. We introduce Stylus, a training-free framework that treats facial style transfer as navigation on the curved latent manifold of a pretrained diffusion model. Rather than interpolating in flat Euclidean space, \method refines the interpolated content–style point by lowering a ratio-weighted discrete geodesic energy under the pullback metric, made tractable by a rank-one power-iteration approximation of the Jacobian. This principle drives two complementary controls: content–style latent interpolation for smooth semantic fusion, and self-attention query optimization for spatially adaptive style injection. The resulting dual-control mechanism provides continuous, fine-grained control over stylization strength while preserving identity-critical facial structure, all without additional training. Experiments show that Stylus achieves the strongest identity and content preservation at competitive stylization strength.
Read the full abstract
Robotics
Discrete Model-Based Diffusion for Combinatorial Optimization in Multi-Robot Task Allocation
Decision making in multi-robot systems often involves discrete decision variables, leading to NP-hard combinatorial optimization problems that must be solved within seconds for near real-time implementation. Denoising diffusion has emerged as a powerful tool for such problems, but existing approaches face a trade-off: Model-Based Diffusion (MBD) incorporates diffusion in sampling-based optimization (SBO) without training but only over continuous spaces, while existing diffusion-based combinatorial solvers operate over discrete spaces but require offline training. To this end, we propose Discrete Model-Based Diffusion (Discrete MBD), a training-free SBO method that directly optimizes discrete variables by approximating categorical reverse transition distributions through Monte Carlo sampling. We establish a finite-step lower bound on the probability of finding a global minimizer and show that this bound converges to one as the number of denoising steps increases. Extensive experiments on generic integer programs and heterogeneous multi-robot coalition formation and scheduling demonstrate that Discrete MBD outperforms SBO baselines in nearly all settings while achieving solution quality competitive with that of mixed-integer programming and reinforcement learning methods at a fraction of their computational cost.
Read the full abstract
Algebraic Geometry
$5I(K_X+\Delta)$ is very ample on stable log surfaces
We prove that Bombieri's bound for canonical models of surfaces of general type extends to stable log surfaces, that is, $mI(K_X+\Delta)$ is very ample for every $m\geq 5$ and every stable log surface $(X,\Delta)$ over $\mathbb{C}$ with reduced boundary and global Cartier index $I$. On the way to this result we determine the cases where $|3I(K_X+\Delta)|$ has base points.
Read the full abstract
Combinatorics
Positive Codegree Density of Suspensions of Graphs
We determine the exact positive codegree density of suspensions of graphs $S^r(G)$ when $\chi(G)=t$ and $t-2$ is prime power. The value is $\gamma^+(S^{r}(G)) = \frac{(t-3)(t-2)^{r-1}}{(t-2)^r-1}$. The case of suspensions of $K_4$ gives the positive codegree density of daisies, answering a question of Balogh, Halfpap, Lidický and Palmer. The lower bound construction is from Ellis, Ivan and Leader, while we establish the corresponding upper bound based on a weighted graph inequality.
Read the full abstract
Robotics
Fastener-centric acquisition and AND/OR graph representation of state-dependent disassembly knowledge from CAD-derived virtual assemblies
Design-stage disassemblability assessment requires knowledge of product structure and the operations available in each assembly state, but CAD assemblies do not directly specify state-dependent disassembly operations. This paper proposes a fastener-centric virtual-disassembly framework that acquires such knowledge from CAD-derived virtual assemblies. Fasteners jointly releasing the same constraint are grouped into operation-level sets, avoiding enumeration of arbitrary individual-fastener orders. Their removability is evaluated through state-dependent visibility and tool-access checks and torque-driven virtual-removal trials. After fastener removal, graph-based reasoning identifies candidate moving units, whose separation is evaluated by initial-motion screening, path generation, and physics-based validation. The acquired operations, states, dependencies, validated paths, and simulation-derived attributes are organized into a Transformed AND/OR Graph for subsequent planning and design evaluation. Exhaustive verification with a controlled toy assembly found no branching violations for introduced tool-accessibility and precedence conditions. Evaluation with a 54-component OpenManipulator Chain gripper demonstrated state-dependent fastener-removal and part- or subassembly-separation operations. Visibility and tool-accessibility evaluations pruned 51.4% of fastener-set candidate evaluations, while removal of obstructing components enabled previously inaccessible fastener operations. Operation-level grouping and screening thereby limited detailed physical validation to mechanically plausible candidates. The results demonstrate the feasibility of acquiring product-specific disassembly knowledge from CAD-derived virtual assemblies before manufacture, supporting subsequent disassembly reasoning and design-for-disassembly evaluation.
Read the full abstract
Machine Learning
Looping Deeper Without Growing the KV Cache: Recurrent Depth for Causal Encoder-Decoders
Looped language models deepen computation without adding parameters, yet caching full-context key-value (KV) states across layers and rounds severely bounds serving batch size. We design a looped causal encoder-decoder (CED) that achieves recurrent depth with zero round-wise KV growth, addressing where and how to loop an asymmetric stack in which only sparse encoder layers generate global KV. We find that recurrence must be placed in the encoder to iteratively refine global KV representations rather than re-reading static decoder projections, which both drives language-modeling gains and preserves long-context scaling. Furthermore, training the encoder loop under stochastic recurrence depth induces a fixed-point equilibrium where every round losslessly shares the last round's KV cache. Compounding cross-layer and cross-round KV sharing, the resulting Looped CED matches an unlooped CED's minimal memory footprint and recovers 86-93% of a looped Transformer's gain with 1/6 of its global KV. To test serving at production scale, we loop the 552B-parameter DeepSeek-V4.1-Flash for 8 rounds on one 4xGB200 node: with per-layer and per-round KV the loop cannot fit a single 128K-token request, whereas Looped CED holds 36.8x less KV per token, fits 38 such requests (as many as the unlooped model), and serves 3.6x higher throughput on 8K-token prompts.
Read the full abstract
Algebraic Geometry
Unimodal Singularities of Parametrized Space Curves
The aim of the article is to classify unimodal parametrized space curve singularities defined over an algebraically closed field of characteristic $0$. We provide explicit normal forms of parametrizations up to $\mathcal A$-equivalence. The parametrizations are given by convergent power series over a real valued field, which includes parametrizations given by formal power series and by, for example, convergent power series over the complex numbers or the field of Puisieux series. The list of normal forms is surprisingly simple.
Read the full abstract
Optimization and Control
Uniform symmetric spectral finiteness of semialgebraic cones
Cone-constrained eigenvalues of a matrix are defined through a complementarity problem constrained by a closed convex cone. We show that every closed semialgebraic convex cone is uniformly symmetrically spectrally finite; that is, the cone spectra, the sets of the cone-constrained eigenvalues, of symmetric matrices relative to the cone are uniformly finite with respect to the matrices. The uniform bound is explicit, and as an application, we derive a uniform bound for the cardinality of spectra of symmetric matrices relative to Löwner cones. In addition, using the uniform finiteness of cone spectra, we also show the uniform finiteness of the number of singular values relative to two closed semialgebraic convex cones. For non-semialgebraic cones, we observe two contrasting phenomena. On the one hand, we show that the exponential cone is uniformly symmetrically spectrally finite. On the other hand, we provide a closed non-semialgebraic convex cone such that the associated cone spectra of symmetric matrices are finite but are not uniformly finite.
Read the full abstract
Metric Geometry
A Topological Approach to the Discrete Dual Minkowski Problem
The discrete dual Minkowski problem asks for the necessary and sufficient conditions on a finite discrete measure $\mu$ on $S^{n-1}$ such that $\mu$ is given by the $q^{th}$ dual curvature measure of a convex polytope in $\mathbb{R}^n$. In this paper, using topological degree theory, a novel framework is developed to prove existence of solutions. This approach is applied to the case of polytopes in $\mathbb{R}^n$ for $q>0$ and $n\geq 2$ whose outer normals are in general position while also subject to a constraint on their distribution on $S^{n-1}$. Moreover, existence of solutions is also proved in the plane under a different constraint on the outer normals by proving that a family of non-linear second-order recurrence relations do not have non-constant $N$-periodic solutions.
Read the full abstract
Materials Science
TASCopilot: A provenance-preserving assistant for triple-axis neutron scattering
Experimental decisions often depend on evidence scattered across papers, instrument logs, data files, and specialist software. A scientific assistant must connect that evidence without turning a plausible interpretation into an unsupported conclusion. TAS Co-pilot addresses this problem for triple-axis spectroscopy (TAS), which uses neutrons to probe material structure and excitations. It preserves provenance: the sources, processing steps, and review decisions behind each result. Independently maintained scientific tools communicate through structured requests and results; validation rules and human review govern how evidence can support a claim. Language models extract information and explain outcomes but do not approve scientific claims or control the instrument. Dependency-aware reduction executes existing scientific routines and reuses unchanged intermediate results. A retrospective SrMn$_2$Sb$_2$ study connects literature on a related material and historical alignment records to modeling of magnetic excitations in 20 measured scans. A live BT-7 study uses a dedicated instrument interface for fine alignment and 40 adaptively selected measurements. Agreement with a conventional measurement of the same crystal at shared conditions supports the acquisition and analysis chain. These cases demonstrate a methodology for maintaining inspectable evidence across literature, numerical analysis, and measurement at a user facility.
Read the full abstract
Quantum Physics
PriorTilt: Decoding quantum LDPC codes with sparse perturbations of belief propagation priors
Although belief propagation (BP) is a widely adopted decoder for classical low-density parity-check (LDPC) codes, its accuracy is limited in the quantum setting. We introduce \PriorTilt{}, a variant of BP obtained by selectively perturbing a small fraction of its prior log-likelihood ratios using offsets determined by recent posterior and outgoing-message statistics, while preserving the same message-passing rules. We compare \PriorTilt{} with Relay-BP and beam search decoders for circuit-level decoding of three benchmark codes. \PriorTilt{} achieves a lower logical error rate (LER) than both baselines at every tested data point. For the $[[144,12,12]]$ Gross code BB144 at noise rate $p=10^{-3}$, the LER reductions reach $4.3\times$ relative to Relay-BP and $3.1\times$ relative to the beam search decoder. At $p=10^{-3}$, \PriorTilt{} has lower mean decoding times than Relay-BP on all three codes and achieves mean decoding speedups of approximately $2.6\times$–$3.0\times$ over the beam search decoder. For BB144 at $p=10^{-3}$, \PriorTilt{} achieves an LER approximately $610\times$ lower than the standard min-sum BP decoder with the same iteration budget, while perturbing fewer than $0.5\%$ of its priors per round on average.