Geometric Topology
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
Thursday 12 March
We define graftable curves on real projective surfaces. In particular, we construct graftable ones in Hitchin case and show that real projective structures with the same Hitchin holonomy, carrying the same weight type, are related to each other via multi-graftings.
Optimization and Control
Families of Two-Impulse Optimal Rendezvous Transfers Between Elliptic Orbits
The classical fuel-optimal two-impulse rendezvous problem between Keplerian orbits is revisited from a family-based perspective. Conventional approaches often yield isolated optimal solutions whose mutual relationships remain unclear; yet, when re-parameterized appropriately, seemingly unrelated optima are revealed to be connected members of continuous solution families. To expose this structure, the proposed framework enforces a subset of first-order necessary optimality conditions and traces the resulting one-parameter families via numerical continuation. The families are classified using Hessian-based criteria and Primer Vector Theory, and are projected onto porkchop plots to connect the angular and temporal domains. Representative case studies reveal the emergence, merging, and disappearance of locally optimal branches under variations in orbital geometry, supplying a global map of the solution landscape. This complementary perspective clarifies the robustness of optimal solutions and identifies alternative near-optimal transfers in the vicinity of a nominal trajectory.
Read the full abstract
Optimization and Control · Cross-listed from quant-ph
Quantum mechanical framework for quantization-based optimization: from Gradient flow to Schroedinger equation
This work presents a quantum mechanical framework for analyzing quantization-based optimization algorithms. The sampling process of the quantization-based search is modeled as a gradient-flow dissipative system, leading to a Hamilton-Jacobi-Bellman (HJB) representation. Through a suitable transformation of the objective function, this formulation yields the Schroedinger equation, which reveals that quantum tunneling enables escape from local minima and guarantees access to the global optimum. By establishing the connection to the Fokker-Planck equation, the framework provides a thermodynamic interpretation of global convergence. Such an analysis between the thermodynamic and the quantum dynamic methodology unifies combinatorial and continuous optimization, and extends naturally to machine learning tasks such as image classification. Numerical experiments demonstrate that quantization-based optimization consistently outperforms conventional algorithms across both combinatorial problems and nonconvex continuous functions.
Read the full abstract
Representation Theory
Tensor Product and the Stable Green Ring of the Symmetric Group Algebra $F\mathfrak{S}_p$
We give an explicit formula for the decomposition of the tensor product of any two indecomposable non-projective modules for the symmetric group algebra $F \mathfrak{S}_p$ modulo projective modules. In particular, we show that the tensor product of two simple modules is semisimple modulo projectives. We also compute the Benson–Symonds invariants for all such indecomposable non-projective modules.
Read the full abstract
Optimization and Control
Simultaneous estimation of multiple discrete unimodal distributions under stochastic order constraints
We study the problem of estimating multiple discrete unimodal distributions, motivated by search behavior analysis on a real-world platform. To incorporate prior knowledge of precedence relations among distributions, we impose stochastic order constraints and formulate the estimation task as a mixed-integer convex quadratic optimization problem. Experiments on both synthetic and real datasets show that the proposed method reduces the Jensen-Shannon divergence by 2.2% on average (up to 6.3%) when the sample size is small, while performing comparably to existing methods when sufficient data are available.
Read the full abstract
Optimization and Control
Blind Hyperspectral and Multispectral Images Fusion: A Unified Tensor Fusion Framework from Coupled Inverse Problem Perspective
Hyperspectral and multispectral images fusion aims at integrating a low-resolution hyperspectral image (LR-HSI) and a high-resolution multispectral image (HR-MSI) to construct a high-resolution hyperspectral image (HR-HSI). It is generally assumed that spatial blurring operator and spectral response operator are prior-known. However, such an assumption is extremely restrictive in practice. To overcome this limitation, this paper formulates blind fusion as a coupled inverse problem, integrating blind deconvolution in the spatial domain with blind unmixing in the spectral domain. From this novel perspective, we propose a unified tensor fusion framework capable of flexible self-adjustment and real-time fusion without pre-training. We further introduce an optimization model for the joint estimation of the target HR-HSI, the spatial point spread function, and the spectral response function. To solve this model, we devise a partially linearized alternating direction method of multipliers (ADMM) algorithm with Moreau envelope smoothing, accompanied by the rigorous convergence analysis. An initialization estimator tailored to the specific characteristics of the fusion problem is proposed. Numerical comparisons with state-of-the-art methods on both synthetic and real-world datasets demonstrate the compelling performance of the proposed method.
Read the full abstract
Group Theory
Modular Cocycles and Haar-Type Measures on Topological Loops
This paper is a continuation of the author's companion work \cite{InoueQuasi} on Haar-type measures for topological quasigroups, where the quasigroup setting was analyzed in connection with Kunen's theorem. We extend that framework to locally compact topological loops and study Haar-type (quasi-invariant) Radon measures together with modular cocycles describing the distortion of such measures under translations. Unlike the classical group case, the composition of translations in a loop is affected by the failure of associativity, which produces an additional correction term governed by an associativity deviation map. We derive the resulting cocycle relation and show that loop identities, in particular Moufang- and Kunen-type identities, impose structural restrictions on the modular data. In the associative limit, the cocycle reduces to the classical modular function of a locally compact group.
Read the full abstract
Information Theory
Strict Optimality of Frequency and Distribution Estimation Under Local Differential Privacy
This paper establishes the strict optimality in precision for frequency and distribution estimation under local differential privacy (LDP). We prove that a linear estimator with a symmetric and extremal configuration, and a constant support size equal to an optimized value, is sufficient to achieve the theoretical lower bound of the $\mathcal{L}_2$ loss for both frequency and distribution estimation. The theoretical $\mathcal{L}_1$ lower bound is also achieved asymptotically. Furthermore, we derive that the communication cost of such an optimal estimator can be as low as $\log_2(\frac{d(d-1)}{2}+1)$ bits, where $d$ denotes the dictionary size, and propose an algorithm to generate this optimal estimator. In addition, we introduce a modified Count-Mean Sketch and demonstrate that it is practically indistinguishable from theoretical optimality with a sufficiently large dictionary size (e.g., $d=100$ for a privacy parameter of $\epsilon = 1$). We compare existing methods with our proposed optimal estimator to provide selection guidelines for practical deployment. Finally, the performance of these estimators is evaluated experimentally, showing that the empirical results are consistent with our theoretical derivations.
Read the full abstract
Analysis of PDEs
A Nash stratification inequality and global regularity for a chemotaxis-fluid system on general 2D domains
Incompressible fluid advection has been shown to facilitate singularity suppression in various differential equations, often by mixing-enhanced diffusion or by dimension-reduction effects. To aid with the study of such scenarios, we prove a refinement of the classical Nash inequality, $$ \|f-f_M\|_{L^2}^2 \lesssim \|f-f_M\|_{L^1}^{8/7}\|\nabla f\|_{L^2}^{6/7} + \|\partial_1 f\|_{\dot{H}_0^{-1}}^{2\vartheta} \|f-f_M\|_{L^1}^{1-\vartheta}\|\nabla f\|_{L^2}^{1-\vartheta}, $$ for $f \in H^1$ with mean $f_M$ over a smooth bounded planar domain under the main constraint of having connected horizontal cross-sections. The first term on the right-hand side follows the classical Nash scaling for a formal dimension of $3/2$. The second term introduces a mixing norm that measures how far $f$ is from being stratified. The proof provides an explicit exponent $0 < \vartheta \ll 1$. As an application, we study the 2D parabolic-elliptic Patlak–Keller–Segel (PKS) chemotaxis model over the aforementioned large class of bounded domains. This aggregation-diffusion equation is well-known to produce finite-time singularity formation for large-mass data in finite domains. Using the above Nash stratification inequality, we prove that the 2D PKS equation becomes globally regular when actively coupled via buoyancy to a fluid obeying Darcy's law for incompressible porous media flows. This result holds for arbitrarily large $C^\infty$ initial data, far from perturbative regimes, and for arbitrarily weak coupling strength. Moreover, the spatial domain can have "bottle neck" regions and boundary segments of large curvature. The argument conceptualizes and generalizes recent work by Hu, Yao, and the first author on the analogous result for the periodic channel.
Read the full abstract
Number Theory
Introduction to Dieudonné modules and supersingular abelian varieties revisited
In this Note we present an expository account for \dieu modules and revisit supersingular abelian varieties. We give a simple proof of the uniqueness of products of two or more supersingular elliptic curves (a theorem due to Deligne, Ogus and Shioda), and of Oort's theorem for superspecial abelian varieties.
Read the full abstract
Commutative Algebra
The generators of a colon ideal with an application to the weak Lefschetz property for monomial almost complete intersections in three variables
Much progress has been made in classifying when the weak Lefschetz property holds for $A=\mathbb{F}[x,y,z]/I$ where $\text{char}(\mathbb{F})=0$ and $I=(x^{d_{1}},y^{d_{2}},z^{d_{3}},x^{a_{1}}y^{a_{2}}z^{a_{3}})$ is a monomial almost complete intersection. We connect this problem to the setting of two variables through a certain relation. In so doing, we are led to determine explicit formulas for the generators of the colon ideal $(x^{d_{1}},y^{d_{2}}):(x+y)^{a_{3}}$. With these generators in hand, we construct a matrix and show that failure of WLP for $A$ is dictated by the vanishing of a certain polynomial (namely the determinant of our matrix) when $A$ is level. We further show in the level case that a conjecture first posed by Migliore, Miró-Roig, and Nagel is true in a few new cases.
Read the full abstract
Algebraic Geometry
Cylinders in weighted Fano varieties
Cylinders in Fano varieties receives a lot of attentions recently from the viewpoints of birational geometry and unipotent geometry. In this article, we provide a survey of several known et new results concerning the anti-canonically polar cylindricity of quasi-smooth, well-formed weighted Fano complete intersections in weighted projective spaces.
Read the full abstract
Optimization and Control · Cross-listed from cs.LG
Slack More, Predict Better: Proximal Relaxation for Probabilistic Latent Variable Model-based Soft Sensors
Nonlinear Probabilistic Latent Variable Models (NPLVMs) are a cornerstone of soft sensor modeling due to their capacity for uncertainty delineation. However, conventional NPLVMs are trained using amortized variational inference, where neural networks parameterize the variational posterior. While facilitating model implementation, this parameterization converts the distributional optimization problem within an infinite-dimensional function space to parameter optimization within a finite-dimensional parameter space, which introduces an approximation error gap, thereby degrading soft sensor modeling accuracy. To alleviate this issue, we introduce KProxNPLVM, a novel NPLVM that pivots to relaxing the objective itself and improving the NPLVM's performance. Specifically, we first prove the approximation error induced by the conventional approach. Based on this, we design the Wasserstein distance as the proximal operator to relax the learning objective, yielding a new variational inference strategy derived from solving this relaxed optimization problem. Based on this foundation, we provide a rigorous derivation of KProxNPLVM's optimization implementation, prove the convergence of our algorithm can finally sidestep the approximation error, and propose the KProxNPLVM by summarizing the abovementioned content. Finally, extensive experiments on synthetic and real-world industrial datasets are conducted to demonstrate the efficacy of the proposed KProxNPLVM.
Read the full abstract
Dynamical Systems
Multiple timescale dynamics of conductance-based models of brainstem locomotor neurons
The pedunculopontine nucleus (PPN) is a heterogeneous brainstem locomotor hub implicated in Parkinson's disease and potentially relevant for its treatment. We propose single-compartment, conductance-based models for three classes of PPN neurons, such that each model reproduces relevant experimentally observed stimulus-dependent responses, including post-inhibitory rebound dynamics, transient low-threshold activity, and gamma band oscillations. To understand the mechanisms underlying these transient responses to current stimulation, we leverage the models' intrinsic multi-timescale structure and apply dynamical system methods designed for multiple timescale systems. By separating fast membrane and channel-gating dynamics from slower gating and calcium processes, we identify specific ionic mechanisms underlying hallmark dynamics across cell types. We also generate new predictions about PPN behavior under a post-inhibitory facilitation protocol.
Read the full abstract
Analysis of PDEs
Regularity thresholds for anomalous dissipation and related phenomena in passive scalars
We prove the absence of anomalous dissipation for passive scalars driven by some random autonomous divergence-free vector fields in $\mathbb T^d$. In dimension $d=2$ we just need continuity almost surely and a mild nondegeneracy condition on the randomness. In dimension $d\geq 3$ we assume a special geometric structure and almost sure Hölder regularity with a Hölder exponent bigger than $\frac{1}{8}$. No regularity is assumed on the passive scalar except for boundedness in the initial data. The proof relies on dimension-theoretic arguments, as opposed to commutator estimates. A consequence of these results is that the same assumptions prevent (almost surely) many other expected properties of turbulent flows, such as anomalous regularization, the Yaglom-Obukhov-Corrsin law, and Richardson diffusion.
Read the full abstract
Statistics Theory · Cross-listed from stat.ME
Bayesian Modular Inference for Copula Models with Potentially Misspecified Marginals
Copula models of multivariate data are popular because they allow separate specification of marginal distributions and the copula function. These components can be treated as inter-related modules in a modified Bayesian inference approach called ”cutting feedback” that is robust to their misspecification. Recent work uses a two module approach, where all $d$ marginals form a single module, to robustify inference for the marginals against copula function misspecification, or vice versa. However, marginals can exhibit differing levels of misspecification, making it attractive to assign each its own module with an individual influence parameter controlling its contribution to a joint semi-modular inference (SMI) posterior. This generalizes existing two module SMI methods, which interpolate between cut and conventional posteriors using a single influence parameter. We develop a novel copula SMI method and select the influence parameters using Bayesian optimization. It provides an efficient continuous relaxation of the discrete optimization problem over $2^d$ cut/uncut configurations. We establish theoretical properties of the resulting semi-modular posterior and demonstrate the approach on simulated and real data. The real data application uses a skew-normal copula model of asymmetric dependence between equity volatility and bond yields, where robustifying copula estimation against marginal misspecification is strongly motivated.
Read the full abstract
Rings and Algebras
Accumulation points of congruence densities of finite lattices
Let $\mathcal W$ be a nontrivial variety of lattices, and let $L$ be a finite lattice in $\mathcal W$. The congruence density of $L$ with respect to $\mathcal W$ is the number of congruences of $L$ divided by the maximum number of congruences of $|L|$-element lattices belonging to $\mathcal W$. We prove that, with respect to the order and multiplication of the real numbers, the set SCD$(\mathcal W)$ of congruence densities of finite members of $\mathcal W$ as well as its topological closure are countably infinite dually well-ordered monoids. We also prove that the set of accumulation points of SCD$(\mathcal W)$ is either a singleton or it is countably infinite; furthermore, it is a singleton if and only if $\mathcal W$ is a subvariety of the variety of modular lattices. This gives a complicated characterization of modularity: a non-singleton lattice $K$ is modular if and only if SCD$(\mathcal V(K))$, where $\mathcal V(K)$ denotes the variety generated by $K$, has only one accumulation point. The class $\mathcal S$ of semimodular lattices is not a variety, but SCD$(\mathcal S)$ is still meaningful; we prove that SCD$(\mathcal S)$ has exactly one accumulation point.
Read the full abstract
Combinatorics
On Directed Graphs with the Same Sum over Arborescence Weights
We show that certain digraphs with the same vertex set but different arc sets have the same sum over the weights of all arborescences with a given root vertex. We relate our results to the Matrix-Tree Theorem and show how they provide a graphical approach for factoring matrix determinants.
Read the full abstract
Complex Variables
Integral mean estimates for $(\alpha,\beta)$-harmonic functions
We establish sharp $L^p$ integral mean estimates for $(\alpha,\beta)$-harmonic functions on the unit disk. Explicit bounds for the functions and their partial derivatives are obtained in terms of boundary data, by means of the associated Poisson-type kernel and hypergeometric function representations. As applications, we derive coefficient estimates and Hardy space-type results, extending well-known inequalities for classical harmonic and $\alpha$-harmonic functions to the $(\alpha,\beta)$-harmonic setting.
Read the full abstract
Probability · Cross-listed from econ.TH
Stochastic Optimization and Coupling
We study optimization problems in which a linear functional is maximized over probability measures that are dominated by a given measure according to an integral stochastic order in an arbitrary dimension. We show that the following four properties are equivalent for any such order: (i) the test function cone is closed under pointwise minimum, (ii) the value function is affine, (iii) the solution correspondence has a convex graph with decomposable extreme points, and (iv) every ordered pair of measures admits an order-preserving coupling. As corollaries, we derive the extreme and exposed point properties involving integral stochastic orders such as multidimensional mean-preserving spreads and stochastic dominance. Applying these results, we generalize Blackwell's theorem by completely characterizing the comparisons of experiments that admit two equivalent descriptions – through instrumental values and through information technologies. We also show that these results immediately yield new insights into information design, mechanism design, and decision theory.
Read the full abstract
Number Theory
On the distribution of shapes of totally real multiquadratic number fields
The shape of a number field $K$ of degree $m$ is defined as the equivalence class of the lattice of integers under linear operations generated by rotations, reflections, and positive scalar dilations. It may be viewed as a point in the space of shapes $\mathscr{S}_{m-1} = \mathrm{GL}_{m-1}(\mathbb Z)\backslash \mathrm{GL}_{m-1}(\mathbb R)/\mathrm{GO}_{m-1}(\mathbb{R})$. The double quotient space is equipped with a natural measure $\mu$ which is induced from the Haar measure on $\mathrm{GL}_{m-1}(\mathbb R)$. We study the distribution of shapes of totally real multiquadratic number fields of degree $m:=2^n$ in which $2$ is unramified. We show that the distribution is governed by the restriction of $\mu$ to a certain torus orbit in $\mathscr{S}_{m-1}$. Our result resolves a conjecture of Haidar.
Read the full abstract
Algebraic Topology
Topological Hochschild homology of truncated Brown-Peterson spectra II
We compute topological Hochschild homology of $\mathbb{E}_3$-MU-algebra forms of the second truncated Brown-Peterson spectrum with Adams summand coefficients at $p=2$ and conditionally at arbitrary primes. We also provide a new computational tool, a variant of the Brun spectral sequence, for computing topological Hochschild homology of truncated Brown-Peterson spectra with certain coefficients. As a consequence, we show that $\mathbb{E}_3$-MU-algebra forms of truncated Brown-Peterson spectra $\mathrm{BP}\langle n\rangle$ are not Thom spectra at the prime $p=2$ for any $n\ge 2$.
Read the full abstract
Analysis of PDEs
Inertial Limit of global weak solutions for Compressible Navier–Stokes
We investigate the inertial limit of the compressible Navier–Stokes system posed on the $3$-dimensional torus, and allowing for regions of vacuum. Considering global-in-time finite-energy weak solutions of a scaled system, we rigorously establish convergence to a limiting system in which the momentum equation reduces to a stationary elliptic balance between pressure and viscous forces. In this limit, the scaled kinetic energy vanishes, reflecting an overdamped regime, and the limiting weak solution satisfies an exact energy equality. Our analysis relies on uniform a priori estimates, renormalized techniques, and compactness arguments in the Lions–Feireisl framework, providing a mathematically rigorous analysis for the overdamped dynamics arising from vanishing inertia in compressible viscous flows.
Read the full abstract
Combinatorics
Structural and Polynomial-Time Results on Core and Corona in Odd-Bicyclic Graphs
Let $\core G$ and $\corona G$ denote the intersection and the union, respectively, of all maximum independent sets of a graph $G$. In this work, we show that for a graph with at most two odd cycles, $\a{\core G}+\a{\corona G}$ is equal to $2\alpha(G)$, $2\alpha(G)+1$, or $2\alpha(G)+2$, and we precisely characterize when each value occurs. We further characterize graphs with at most two odd cycles that admit the core–corona partition $V(G)=\corona G\ud N(\core G)$, extending known results for König–Egerváry and almost bipartite graphs. Deciding whether $\core G=\emptyset$ is known to be NP-hard. As an algorithmic consequence of the obtained results, we show that the core, independence number and the corona can be computed in polynomial time for this class of graphs.
Read the full abstract
Combinatorics
Core and Corona in 2-Bicritical Odd-Bicyclic Graphs
Let $\core G$ and $\corona G$ denote the intersection and the union, respectively, of all maximum independent sets of a graph $G$. A graph is called \emph{$2$-bicritical} if $\a{N(S)}>\a S$ for every nonempty independent set $S$. Pulleyblank 1979 showed that almost all graphs are $2$-bicritical. In this paper, we study the structure of maximum independent sets in $2$-bicritical graphs with at most two odd cycles. Using ear–pendant decompositions, we obtain a complete structural classification of these graphs into four families: one-odd cycle, fused-odd, even-linked, and odd-linked graphs. For each family, we compute explicitly $\alpha(G)$, $\core G$, and $\corona G$, and describe the corresponding matching structure. We prove that $\a{\core G}+\a{\corona G}$ equals either $2\alpha(G),2\alpha(G)+1$ or $2\alpha(G)+2$, and we give a complete, purely structural characterization of the graphs in each case in terms of the relative position of their odd cycles. These results extend a theory originally developed for König–Egerváry graphs and later for almost bipartite graphs to a broader non-König–Egerváry setting.
Read the full abstract
Combinatorics
A characterization of graphs with $\a{\corona G}+\a{\core G}=2\alpha(G)+1$
A Kőnig–Egerváry graph is a graph $G$ satisfying $\alpha(G)+\mu(G)=n(G)$, where $\alpha(G)$, $\mu(G)$, and $n(G)$ denote the independence number, the matching number, and the order of $G$, respectively. Let $\textnormal{core}(G)$ and $\textnormal{corona}(G)$ be the intersection and the union of all maximum independent sets of $G$. In this paper, we provide a complete characterization of graphs satisfying $\a{\corona G}+\a{\core G}=2\alpha(G)+1$, thus giving a solution to an open problem posed by Levit and Mandrescu. It is known that for a non-Kőnig–Egerváry graph with a unique odd cycle, the following hold: $\ker G=\textnormal{core}(G),\allowbreak\ \left|\textnormal{corona}(G)\right| +\left|\textnormal{core}(G)\right| =2\alpha(G)+1,\allowbreak\ \textnormal{corona}(G)\cup N(\textnormal{core}(G))=V(G)$. We extend these three results to a family of graphs containing an arbitrarily large number of odd cycles.
Read the full abstract
Representation Theory
Exceptional theta correspondences via Plancherel formulas for rank one symmetric spaces
We consider the minimal representation of (a finite cover of) the conformal group of a simple split Jordan algebra over $\mathbb{R}$ or $\mathbb{C}$, whenever it exists. The conformal group contains a natural dual pair $G\times G'$, where $G$ is essentially the automorphism group of the Jordan algebra and $G'$ is either $\operatorname{PSL}(2,\mathbb{R})$, $\operatorname{PGL}(2,\mathbb{R})$ or $\operatorname{PGL}(2,\mathbb{C})$. The groups $G$ that arise in this way include the complex exceptional group of type $F_4$ as well as its compact and split real form. We explicitly determine the direct integral decomposition of the minimal representation restricted to the corresponding cover of $G\times G'$. This yields a one-to-one correspondence between certain representations of $G$ and (a finite cover of) $G'$. The representations of $G$ that occur in this correspondence are in the support of the Plancherel measure for a rank one symmetric space for $G$, and the proof makes use of the corresponding Plancherel formula.
Read the full abstract
Dynamical Systems
Framing local structural identifiability and observability in terms of parameter-state symmetries
We introduce a subclass of Lie symmetries, called parameter-state symmetries, to analyse the local structural identifiability and observability of mechanistic models consisting of state-dependent ODEs with observed outputs. These symmetries act on parameters and states while preserving observed outputs at every time point. We prove that locally structurally identifiable parameter combinations and locally structurally observable states correspond to universal invariants of all parameter-state symmetries of a given model. We illustrate the framework on four previously studied mechanistic models, confirming known identifiability results and revealing novel insights into which states are observable, providing a unified symmetry-based approach for analysing structural properties of dynamical systems.
Read the full abstract
Wednesday 11 March
Optimization and Control
The Keplerian Traveling Salesperson Problem
We address a fundamental challenge in space mission design and space logistics: planning interplanetary trajectories for missions that must rendezvous with multiple bodies. Such mission occur, for instance, in active debris removal, in-orbit servicing, or asteroid belt exploration. We model these problems as a variant of the Traveling salesperson problem (TSP), which we term the Keplerian TSP (KTSP). Unlike the well-studied TSP, the KTSP accounts for the motion of orbital targets, leading to time-dependent and asymmetric transfer costs that capture key real-world effects in astrodynamics. We provide a rigorous formalization of the KTSP and release a benchmark suite to support its study. Central to our approach is a time-unfolding technique that reformulates the continuous problem as a discrete optimization task in a time-expanded network. This representation makes the benchmark accessible to researchers in discrete optimization even without prior knowledge of celestial mechanics. We also develop an alternative encoding as an integer linear program using Interval-based Dynamic Discretization Discovery to handle the time-dependent nature of transfers. We leverage state-of-the-art ILP solvers to solve the KTSP instances, accompanied by a detailed computational study that highlights their strengths and limitations. We complement these exact methods with an initial solution heuristic, an improvement heuristic, and preprocessing routines that preserve optimality.
Read the full abstract
Information Theory · Cross-listed from cs.CL
The Stepwise Informativeness Assumption: When Do Entropy Dynamics Track Reasoning Correctness in LLMs?
Recent work uses entropy-based signals at multiple representation levels to study reasoning in large language models, but the field remains largely empirical. From a theoretical perspective, reductions in model uncertainty do not necessarily reflect progress toward correctness. In this work, we characterize when such alignment can be expected through the Stepwise Informativeness Assumption (SIA), which requires entropy-reducing reasoning-prefix extensions increase the probability assigned to valid answers in expectation. We show that next-token prediction can transfer answer-relevant information from training data to the model, but does not, by itself, guarantee alignment with correctness. Instead, we propose that this alignment depends on how the training regime couples reasoning traces to correct answers. Furthermore, we hypothesize observable signatures of this phenomenon and characterize its expected failure regimes. We empirically investigate how entropy dynamics relate to correctness across four reasoning benchmarks (GSM8K, ARC, SVAMP, MATH) and diverse open-weight language models spanning base, instruction-tuned, and RL-trained regimes. Our results show that entropy-correctness alignment strengthens with post-training but can weaken when task difficulty exceeds model capabilities.