Analysis of PDEs
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Saturday 12 September
In the one-dimensional log-modified nonlinear Schrödinger equation, there are finitely many normalised standing-waves, up to space translation and phase multiplication. In the defocusing case, there is exactly one normalised standing-wave having a prescribed mass and least energy. Both results are obtained by exploiting properties of the mass function: the analyticity in the focusing case, and the monotonicity in the defocusing case.
Information Theory
Adaptive Coding for Two-Way Wiretap Channels with One-Sided Strong Secrecy
We study a discrete memoryless two-way wiretap channel under strong one-sided secrecy, where User 1's message is confidential and User 2's message is not required to be secret. We develop non-adaptive and key-assisted adaptive inner bounds from a common resolvability estimate for selected index components. For the adaptive construction, we prove a multiround leakage theorem that accounts both for information revealed about a key before its later use and for the possibility that an encoder uses an incorrectly decoded key. The proof includes an ideal-to-implemented coupling, deterministic codebook selection, and a growing-round argument that removes the initialization-rate loss. We eliminate the internal component rates exactly and characterize the adaptive gain at each fixed auxiliary distribution. For an open two-parameter family of binary channels, we also prove a converse for arbitrary non-adaptive stochastic codes and obtain an operational adaptive-versus-non-adaptive separation. For one representative channel, the complete non-adaptive strong one-sided secrecy capacity region is identified exactly.
Read the full abstract
Combinatorics
A Bound Below 2.8 for Tuza's Conjecture
Let $\nu(G)$ be the maximum number of edge-disjoint triangles in a graph $G$ and $\tau(G)$ the minimum number of edges meeting every triangle. Tuza conjectured that $\tau(G)\le 2\nu(G)$. We prove that $\tau(G)\le (165/59)\nu(G)$. The constant $165/59\approx 2.797$ improves the bound $66/23\approx 2.870$ that Haxell proved in 1999. The key observation is that, for a suitable red-blue coloring, the families left over in Haxell's construction contain every triangle with exactly one red edge. Such a family $\mathcal{F}$ admits an exchange that forces certain red edges to lie in a single triangle once the blue edges of a maximum packing are deleted, which gives $\tau(\mathcal{F})\le (8/3)\nu(\mathcal{F})$.
Read the full abstract
Computer Vision
DiVA: Enabling Interactive Digital Life Simulation via Video Models
We present DiVA, a deeply interactive digital life simulator pioneering a new paradigm for long-term, open-ended interactive experiences within digital character worlds. DiVA's architecture pairs a Multimodal Large Language Model (MLLM) as a router with a meticulously designed stacked video pipeline for seamless, multi-turn interactions with action and audio response. To maintain continuity and avoid degradation, we model generation as a three-part coupled system: waiting video, action video, and the transitions between them. These transitions are critically handled by our Anchored Video Continuation (AVC) module, which returns the character to stable states to prevent degradation. By encoding information from the preceding action video segment, AVC ensures smooth transitions, significantly reducing camera jitter and inconsistencies common in current video transition methods. This design also enables complex pose changes (e.g., sitting to standing) typically difficult for audio-driven models. These system designs together ensure high-fidelity identity, coherence, and dynamics for extended experiences. To validate our pipeline design, we comprehensively compare our system against alternatives by replacing our core generation module with mainstream long-video, continuation, and interpolation methods. We further analyze the necessity of the three-stage design, anchor-state selection, transition naturalness, spatial grounding, and the quality-latency trade-off, and we expand the comparison to additional long-form audio-driven avatar models. Results confirm DiVA is markedly superior in maintaining long-term visual quality and realism, validating its effectiveness as a sustainable, interactive simulation.
Read the full abstract
Differential Geometry
Alexandrov–Fenchel inequalities for convex domains in the sphere
We prove the Alexandrov–Fenchel inequalities between any two spherical quermassintegrals for smooth weakly convex domains contained in an open hemisphere, with equality if and only if the domain is a geodesic ball. For strictly convex hypersurfaces, we introduce a globally constrained curvature flow which preserves one quermassintegral and decreases the next one. We establish uniform curvature estimates by combining a pinching estimate, a support function argument, and spherical polarity. As a consequence, the flow exists for all time and converges smoothly and exponentially to a geodesic sphere. The monotonicity of the quermassintegrals gives the full family of Alexandrov–Fenchel inequalities. A short-time mean curvature flow approximation and a localized rigidity argument extend the result, including the equality characterization, to weakly convex domains.
Read the full abstract
Probability
Viscosity Solutions for Singular HJB Equations: BSDE Representations and Stochastic Control
We introduce a notion of viscosity solution for Hamilton–Jacobi–Bellman (HJB) equations with distributional drift, based on paracontrolled test functions and related through a Zvonkin transformation to classical viscosity theory. The equations considered are of the form \[ \left(\partial_t+\frac12\Delta+b\cdot\nabla\right)h(t,x) =-H(t,x,h(t,x),\nabla h(t,x)), \] where $b$ is singular in the sense of \cite{paradistrib} and has regularity $\mathcal C^{-\alpha}$ for $\alpha\in(1/2,2/3)$. Using doubling-of-variables arguments, we derive a priori gradient estimates that also cover Hamiltonians with slightly superquadratic growth in $\nabla h$. We also obtain probabilistic representations through singular control problems for convex $H$ and weak singular forward–backward SDEs for possibly nonconvex Hamiltonians with at most quadratic growth.
Read the full abstract
General Topology
One-Point Metrizable Coarsenings: Gauges and Local Metric Preservation
Let $(X,\tau)$ be metrizable and let $a\in X$. We give a constructive account of metrizable topologies $\sigma\subseteq\tau$ that agree with $\tau$ on $X\setminus\{a\}$. Applying Hausdorff's classical metric collapse construction, for every noncompact $(X,\tau)$ and every compatible metric $d$ we obtain a strict coarsening with a metric $p\le d$ that agrees with $d$ on a common neighborhood of each point other than $a$. A prescribed countably infinite closed discrete set $\{x_n:n\in\N\}\subseteq X\setminus\{a\}$ can be made to satisfy $p(a,x_n)\le\lambda_n$ for any positive null sequence $(\lambda_n)$. The resulting metric is greatest among the metrics dominated by $d$ that satisfy these bounds, and is complete whenever $d$ is complete. We exhibit its realization as a classical metric quotient. We also represent all localized metrizable coarsenings by continuous scalar gauges using a standard cone metric. Inclusion is expressed by the cofinal comparison of sublevel sets familiar from extension-trace theory, while pointwise maximum and minimum realize finite joins and meets. A closed-discrete criterion detects strictness. Standard preservation results for Borel structure, complete metrizability, and Polishness, together with function-space and local-field examples, complete the account.
Read the full abstract
Materials Science
B-type Quadratic Planar Hall Effect as a Probe of Altermagnetic Order
Conventional transport and optical responses, such as the Hall effect, can only distinguish material phases with substantially different symmetries. However, many quantum materials, such as candidate altermagnetic materials, host competing phases with similar structures, making it difficult to differentiate them using conventional transport or optical measurements. Here, guided by symmetry principles, we introduce a B-type quadratic planar Hall effect (B2-PHE) as a sensitive probe of magnetic structures. B2-PHE features a planar Hall current that scales quadratically with the applied magnetic field. This response originates from the interplay among external magnetic fields, intrinsic magnetic order, and quantum geometric properties of the electronic bands. We investigate B2-PHE in two altermagnetic candidates, KV2Se2O and RuO2, whose ground-state magnetic orders are under intense debate. Remarkably, B2-PHE emerges in their altermagnetic phases but is symmetry-forbidden in their antiferromagnetic (nonmagnetic) phases. We also show that B2-PHE should be detectable under a moderate magnetic field on the order of 1 Tesla. These findings establish B2-PHE as an experimentally convenient and potentially unambiguous probe of altermagnetic order, which is otherwise difficult to realize.
Read the full abstract
Artificial Intelligence
ViperQ: Order Flow Pattern Recognition via Auction Market Theory for Reinforcement Learning Trading
Reinforcement learning trading systems published in the academic literature overwhelmingly rely on price-aggregate state representations (OHLCV bars) or limit-order-book depth features, leaving microstructure pattern theories from the practitioner literature, namely Auction Market Theory and Market Profile, without a peer-reviewed computational instantiation. We present ViperQ, a reinforcement learning system whose state representation is built explicitly from Auction Market Theory primitives: Volume Point of Control, Value Area position, Low Volume Node flags, Cumulative Volume Delta divergence, and tape-velocity signatures, assembled into a 20-dimensional Z-normalised vector. Two Proximal Policy Optimisation agents are trained with a prospect theory-grounded asymmetric reward function that penalises losing holds at a magnitude consistent with Kahneman and Tversky's loss-aversion coefficient. Evaluated on a held-out twelve-month partition of institutional tick data the agents have never seen, ViperQ achieves +163.6% ROI on TSLA (-27.5% max drawdown, 27,019 trades) and +116.5% ROI on NVDA (-47.8% max drawdown, 12,892 trades) under zero leverage. The results establish Auction Market Theory features as a tractable structured input modality for sequential decision-making on financial time series and motivate further work on microstructure-aware policy learning.
Read the full abstract
Computation and Language
When Consistency Does Not Mean Reliability: Evaluating Local LLM Judges Against Human Ratings
Large language models (LLMs) are increasingly used to evaluate the responses of other language models. This approach, known as LLM-as-a-Judge, is faster and cheaper than human evaluation. However, a judge may produce consistent scores without necessarily agreeing with human evaluators. In this work, we study this issue using two local open-weight LLM judges, LLaMA-3-8B and Qwen2.5-7B. We evaluate 300 responses generated by an instruction-tuned GPT-2 (124M) model for 100 questions covering five categories: factual knowledge, instruction following, mathematics, reasoning, and writing. Each response is scored by nine human annotators and is evaluated three times by each LLM judge using the same rubric. We compare the judge scores with the average human scores using Pearson correlation, Spearman correlation, mean absolute error (MAE), signed bias, and self-consistency. LLaMA-3-8B shows a Pearson correlation of 0.275 with human scores, while Qwen2.5-7B achieves 0.340. Their MAEs are 27.71 and 18.64, respectively. Despite this limited agreement, both judges show high self-consistency, with exact consistency rates of 97.3% for LLaMA-3-8B and 92.3% for Qwen2.5-7B. These results show that high self-consistency does not necessarily indicate high agreement with human judgments. Our findings highlight the need to evaluate both consistency and human alignment when using local LLMs as automatic judges.
Read the full abstract
Computer Vision
Semantic Privacy Protection with Utility Preservation for 3D Point Clouds
Point cloud data face serious semantic privacy risks during acquisition, transmission, and cross-institutional sharing. Existing methods mostly rely on geometric perturbation or destructive encryption, which can reduce the recognizability of the original class but often impair downstream usability. This paper proposes a class-transfer-based semantic encryption framework for point clouds, aiming to conceal original class information while preserving task utility and supporting authorized recovery. Specifically, we construct a unified latent space with a shared-backbone Normalizing Flow, and combine LoRA and FiLM to achieve parameter-efficient class-conditional adaptation. We further introduce diffusion-guided flow alignment to regularize the latent distribution, construct an energy-based category transition graph, and obtain an optimal class-transfer table through global matching. Then, a latent-space Neural ODE continuously evolves source-class latents into target-class latents, which are decoded into target-class point clouds through the inverse flow. We adopt attacker-oriented metrics, including New-Class Recognition Rate (NCRR), Original-Class Leakage Rate (OCLR), and Original Label Recovery Rate (OLRR), to evaluate privacy and utility. Experiments on classification and segmentation benchmarks show that the proposed method achieves controllable semantic transformation, effectively reduces original-class semantic leakage, preserves downstream learnability in the protected domain, and supports reliable authorized reconstruction.
Read the full abstract
Numerical Analysis
Domain splitting with localized implicit prediction for wave equations
We propose and analyze a non-iterative domain decomposition time integrator for the linear acoustic wave equation. The method is based on an overlapping decomposition in space and uses a localized implicit prediction to provide artificial boundary data at the interfaces. We use conforming finite elements of arbitrary order to discretize in space and allow spatially varying material coefficients. We prove an error bound relative to the global Crank–Nicolson approximation that decays exponentially with the size of both the subdomain and prediction overlaps. Compared to earlier work, this new method avoids a CFL-type step size restriction and allows the use of finite elements of arbitrary order. The analysis combines a discrete Saint-Venant principle with new decay and truncation estimates for the localized prediction. Numerical experiments support the analysis and demonstrate stable behavior for all tested time step sizes, accuracy comparable to that of the global Crank–Nicolson approximation, and the potential for efficient parallel execution.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Consistent treatment of flavour-specific neutrino self-interactions in cosmology
We study the impact of flavour-specific neutrino self-interactions on the CMB anisotropy spectrum and consistently take into account neutrino mixing. Starting from the neutrino quantum kinetic equations, we derive the neutrino Boltzmann hierarchies for the neutrino mass states. Our work shows that most scenarios of flavour-specific neutrino self-interactions are well approximated by a combined fluid approach. In practice, this implies that existing CMB bounds on the effective neutrino coupling are weakened by a factor $\sqrt{2}$ for scenarios in which only two neutrino flavours interact (irrespective of which flavours). If only one neutrino flavour interacts the bounds are weaker by a factor $\sqrt{6}$, with the only caveat that the combined fluid approach is less reliable for only $\nu_e$ interacting.
Read the full abstract
Analysis of PDEs
Rigidity of positive rupture solutions to a biharmonic equation with critical negative exponent
We establish two rigidity theorems for positive rupture solutions of the conformally invariant equation $\Delta^2 u=u^{-7}$ in $\mathbb R^3\setminus\{0\}$, which extend continuously to the origin with $u(0)=0$. First, we prove that if the associated conformal metric $g=u^{-4}|dx|^2$ has nonnegative scalar curvature, then every such solution is radially symmetric and has the sharp rupture profile $u(x)\sim(\frac{4}{3})^{\frac{1}{4}}|x|^{\frac{1}{2}}$ as $x\to 0$; in particular, the metric is complete at the origin. The principal novelty is a global rigidity theorem requiring neither curvature nor symmetry: the single global condition $u(x)=o(|x|)$ at infinity forces $u(x)\equiv(\frac{4}{3})^{\frac{1}{4}}|x|^{\frac{1}{2}}$. This result provides a sharp answer to the uniqueness question posed by McKenna and Reichel in a class with no a priori symmetry assumption. Finally, we construct complementary examples demonstrating the essential roles of the curvature and growth hypotheses.
Read the full abstract
Machine Learning
Benchmarking Optimizers to Solve Inverse Problems with Differentiable Physics Simulators
Solving inverse problems with differentiable physics simulators holds the potential to revolutionize scientific discovery and engineering design, as it enjoys both the strict physical correctness from rigorous numerical physics simulators, and the high efficiency and effectiveness from automatic differentiation and gradient-based optimization. However, currently, this paradigm faces performance issues in optimization. In this work, we target benchmarking the performance of different optimizers to solve various inverse problems. We construct 12 differentiable physics simulators spanning physics domains including discrete mechanics, continuous mechanics, atomistic simulations, rendering, and semi-empirical physics models. Based on these simulators, we design corresponding inverse problems that can be categorized into parameter identification, inverse design, and optimal control. Finally, we conduct extensive experiments to compare the performance of different optimizers, including regular first-order methods, approximate second-order methods, as well as global optimizers, on these inverse problems, and analyze the results to provide insights on how to choose and design optimizers for differentiable programming. We hope such benchmarks can inspire the development of more effective optimizers, and further promote the applications of differentiable programming in various scientific and engineering domains.
Read the full abstract
Combinatorics
Extremal function for rooted $K_5$ minors
We show that if an n-vertex 5-connected graph has at least 4n-10 edges, then for any choice of five of its vertices, we can contract disjoint connected subgraphs containing these vertices to obtain $K_5$ as a minor. The bound on the number of edges is the best possible.
Dynamical Systems
Relative mean sensitivity and relative entropy under group actions
In this paper we introduce relative mean sensitivity both in topological and measure-theoretical settings under group actions. It turns out that relative positive entropy is relative mean $n$-sensitive between minimal systems; relative positive $\mu$-entropy for an ergodic measure $\mu$ is relative mean $\mu$-$n$-sensitive.
Read the full abstract
Cryptography and Security
Exploring Automated Vulnerability Identification in JavaScript Code Using Large Language Models
JavaScript powers approximately 98.8% of all websites, making vulnerabilities in its code a significant security risk, yet existing detection approaches such as Static Application Security Testing (SAST) tools often fail to identify many real-world vulnerabilities when applied to isolated code snippets. This paper presents an empirical study of Large Language Model (LLM)-based vulnerability identification for JavaScript programs, evaluating three LLM families (Gemini 1.5 Flash, GPT-4o Mini, DeepSeek-R1-Distill-Llama-8B) across multiple prompting strategies (zero-shot, chain-of-thought, few-shot) and fine-tuning approaches on a dataset of 1,125 JavaScript code snippets spanning five Common Weakness Enumeration (CWE) categories: Injection (CWE-74), OS Command Injection (CWE-78), Cross-Site Scripting (CWE-79), SQL Injection (CWE-89), and Uncontrolled Resource Consumption (CWE-400). Our experiments show that LLMs substantially outperform traditional SAST tools on snippet-level vulnerability identification, with a fine-tuned Gemini 1.5 Flash model achieving 60% detection accuracy compared to near-zero performance from rule-based analyzers. We find that fine-tuning improves accuracy from 29% to 60%, Chain-of-Thought prompting benefits reasoning-capable models such as GPT-4o Mini, few-shot prompting is effective for polymorphic vulnerabilities such as Cross-Site Scripting, and performance varies across vulnerability categories, reaching up to 84% accuracy for structured vulnerabilities such as SQL Injection. These results indicate that LLMs provide a practical approach for automated vulnerability identification in JavaScript code, particularly when combined with task-aligned supervision, though they should complement rather than replace existing security analysis workflows due to limited recall and uneven performance across vulnerability types.
Read the full abstract
Computer Vision
Beyond OCR Accuracy: Text-Centric VQA Under Image Degradation with Modular and End-to-End
Text-centric Visual Question Answering (VQA) requires reading and reasoning over text embedded in images, a task made substantially harder when images suffer from real-world degradation such as motion blur, low resolution, or compression artifacts. While modular OCR-based pipelines and end-to-end vision-language models are both widely used for this task, their comparative robustness under degraded conditions remains underexplored. We present an empirical study comparing two modular pipelines with SA-DBNet, a custom detector architecture combining ResNet-18 with self-attention spatial modeling and deformable convolutions against an end-to-end vision-language baseline, evaluated on 4013 degraded images with 7000 question-answer pairs. Fine-tuned modular pipelines achieve up to 57.50% exact-match accuracy versus 38.00% for the end-to-end baseline, with domain-specific fine-tuning yielding a gain of up to 29.50 percentage points. Critically, we find that conventional OCR error metrics like Character Error Rate and Word Error Rate are unreliable predictors of downstream VQA performance, as semantic reasoning can compensate for recognition failures when contextual cues are present. These findings highlight the importance of task-aware evaluation for text-centric VQA systems under realistic visual conditions. Codes are available here
Read the full abstract
Computer Vision
Realtime-Venus: A full-duplex interaction system with asynchronous delegation
Natural interaction in digital and physical environments requires continuous perception and timely responses. Spoken dialogue relies on acoustic and linguistic cues, while video interaction also requires grounding the conversation in evolving visual context. We present Realtime-Venus, a proactive full-duplex interaction system with two separately trained 9B models: Realtime-Venus-Omni for audio-visual interaction and Realtime-Venus-Audio for spoken interaction. Each model serves as a complete conversational frontend, integrating continuous perception, conversational control, and native speech generation through a shared causal timeline for user inputs, model outputs, and delegation events. A dual-loop runtime coordinates live interaction with background reasoning and tool execution. Foreground interaction continues while Realtime-Venus-Harness executes tasks asynchronously and returns results for integration into the ongoing dialogue. Both models follow a common post-training recipe combining offline understanding, proactive full-duplex trajectories, and delegation workflows. Among the evaluated online models, Realtime-Venus-Omni achieves the highest scores on six of eight video benchmarks, including StreamingBench (70.2%), OVO-Bench (64.7%), and Daily-Omni (81.3%). Across eight audio understanding and spoken question answering benchmarks, Realtime-Venus-Audio leads the compared models on MMAU (78.0%), MMAU-Pro (63.2%), Llama Questions (83.8%), and Speech CMMLU (67.8%), while matching the best VoiceBench AlpacaEval score of 4.81. On Full-Duplex-Bench v1.5, Realtime-Venus-Audio responds to 75% of user interruptions and achieves continuation rates of 97%, 88%, and 86% under backchannels, other-directed speech, and background speech, respectively, exceeding Gemini 3.1 Live and GPT-4o on all three continuation metrics.
Read the full abstract
Optimization and Control
Interior-point proximal methods for nonsmooth optimization in Hilbert spaces with cone-ordered constraints
We study an inexact interior-point method for nonsmooth, possibly nonconvex optimization in a Hilbert space with inequality constraints ordered by a cone in a Banach lattice, with particular emphasis on infinite-dimensional state-constrained optimal control. The objective function is given by the sum of a smooth, possibly nonconvex term and a convex, possibly nonsmooth term with a computable proximal mapping. The constraints are formulated by means of an order cone in a Banach lattice. This setting covers finite-dimensional nonsmooth nonlinear problems with componentwise constraints as well as infinite-dimensional PDE-constrained optimization problems with pointwise state constraints. The method is based on barrier-regularized subproblems, which are solved inexactly by a proximal-gradient method. We consider logarithmic and power-type barriers and derive the differentiability and curvature estimates needed for the convergence and complexity analysis. Under suitable constraint qualifications and compactness assumptions, we establish approximate KKT conditions for the original problem and convergence of the inexact interior-point sequence. For logarithmic and power-type barriers, we derive complementarity estimates and outer iteration bounds; the corresponding power-barrier rates and total inner-outer complexity bounds are stated under explicit barrier-path and uniform smoothness assumptions. For convex problems, we obtain stronger convergence results. We apply the framework to state-constrained semilinear elliptic optimal control and sparse dictionary learning with nonlinear side constraints. Numerical experiments illustrate the proposed method.
Read the full abstract
Robotics
GeomVLA: Unifying Scene, Motion, and Action in 3D
We present GeomVLA, a Vision-Language-Action (VLA) model that unifies perception, latent scene motion prediction, and action generation within a shared robot-centric 3D coordinate frame. Our approach lifts pretrained VLM features into spatially grounded 3D scene tokens using depth and camera calibration, while retaining the semantic representations learned during VLM pretraining. We further introduce a 3D Scene Trajectory Denoiser, a task-conditioned module that learns a latent representation of how scene points are expected to move in 3D. Rather than executing the predicted trajectory as an open-loop plan, GeomVLA extracts intermediate motion tokens from the trajectory denoiser and uses them to condition a 3D flow-based action denoiser through geometry-aware attention. GeomVLA achieves state-of-the-art performance on CALVIN, competitive performance on LIBERO and RoboTwin2.0, and outperforms strong baselines in real-world manipulation settings without robot-action pretraining. Extensive ablations show that future-motion reasoning alone is insufficient: the primary gains are associated with maintaining geometric consistency among scene representation, motion prediction, and robot actions throughout the perception-to-action pipeline.
Read the full abstract
Dynamical Systems
Characterizing resonant solitary states as relative equilibria via normal form reduction
We investigate solitary states in coupled oscillator networks, where a single oscillator detaches from a synchronized cluster and evolves at a distinct mean frequency. We focus on resonant solitary states, in which the mean frequency of the detached oscillator is shaped by resonance with a mode of the cluster. Here, we introduce a normal-form and symmetry-based approach to identify such states. We reduce the high-dimensional network dynamics to a two-dimensional second-order system that couples a rotator representing the solitary oscillator to a harmonic oscillator representing the dominant Laplacian mode of the cluster. We use normal-form transformations to introduce symmetry into this system. Using this symmetry, we reformulate the frequency selection problem as the search for relative equilibria of the normal form symmetry. We derive explicit algebraic conditions for the existence of these relative equilibria, and hence for resonant solitary states and their effective frequencies.
Read the full abstract
Information Retrieval
Odds-Shift Slippage in One-vs-Rest Rankers: Diagnosing and Repairing Reweighting-Induced Top-K Errors
One-vs-rest rankers that show each user the top-$K$ of many rare labels usually counter imbalance with a per-label positive-class weight, scale_pos_weight $= n_-/n_+$. Elkan's identity says such a weight shifts label $j$'s log-odds by $\ln w_j$, so the model ranks by weighted odds rather than by the marginal that is Bayes-optimal for precision@$K$, and suggests inverting the shift afterwards; what a finite learner does with a weight in the thousands, and which repair then works, has not been measured. We call the gap between the promised and the realized shift odds-shift slippage and measure it on matched pairs of LightGBM and MLP models that differ only in the weights. On Santander the weight takes MAP@7 from 0.808 to 0.117; for the boosted pairs the ideal odds shift accounts for 23% of that loss (32% on Instacart; 98% for an MLP pair on the same rows) and slippage for the rest. We prove that a booster whose leaf steps are capped at $c$ realizes at most $T\eta c$ nat of shift in $T$ rounds at rate $\eta$, which a cap sweep confirms, and show that without a cap saturated cells tie at exactly 1.0, beyond the reach of any separable map. The analytic inversion therefore pays only where the shift was realized and nothing saturated, whereas per-label isotonic regression returns the Santander model to 0.784 (0.780 with the calibrator fitted on the validation period), but only if labels without calibration positives are mapped to their prior rather than passed through. On 11 public MULAN benchmarks and 5 learners the weighted model loses more than half of its MAP@$K$ in 8 of 55 cells, and on delicious and Corel5k the same repair returns it to the unweighted level; per-label calibration hurts where positives are scarce, a harm that a cross-validated rule removes. The recipe is released as oddslip.
Read the full abstract
Category Theory
Concept lattices of $Q$-relations via diagonal and back-diagonal categories
Let $Q$ be a commutative unital quantale and $\QRel$ the ordered category of sets and $Q$-relations. We develop a categorical foundation for formal and property-oriented concept lattices of $Q$-relations via the diagonal category $\DQRel$ and back-diagonal category $\BQRel$. Our main results are: (1) the formal concept lattice construction is induced by a representable functor on $\BQRel$; (2) the property-oriented concept lattice construction corresponds to a representable functor on $\DQRel$; and (3) for regular $Q$-relations, the property-oriented concept lattice of their pointwise tensor product is naturally isomorphic to the tensor product of the respective property-oriented concept lattices.
Read the full abstract
Computation and Language
DARE: Dialectical Agentic Reasoning for Structured Knowledge Fact Checking
Structured knowledge fact checking aims to determine the truthfulness of natural language claims by reasoning over structured evidence. Recent program-generation approaches leverage large language models (LLMs) to generate executable graph reasoning programs, achieving strong performance on structured knowledge fact checking benchmarks. However, these methods remain limited by invalid relation generation, single-path reasoning that lacks self-correction, and biased evidence assessment that tends to overestimate supporting signals. We propose Dialectical Agentic Reasoning (DARE), a multi-agent framework that formulates structured knowledge fact checking as an iterative retrieve-reason-reflect process. DARE integrates relation-grounded evidence retrieval to constrain reasoning to valid structures, dialectical bidirectional verification to evaluate evidence from both supporting and refuting perspectives, and confidence-driven meta-reflection to dynamically determine whether additional evidence exploration is necessary. Extensive experiments demonstrate the effectiveness of DARE in structured knowledge fact checking, with an 8B backbone achieving 88.12% accuracy and matching or surpassing GPT-4o-based program-generation baselines, which attests to the efficacy of dialectical agentic reasoning in eliciting the latent reasoning capabilities of LLMs.
Read the full abstract
Artificial Intelligence
Bypass Observation: A Conceptual Design of a Non-Intrusive Layer-Wise Semantic Extraction Architecture
Large language models reason in high-dimensional hidden-state spaces, while users observe only final outputs. We introduce Bypass Observation, a non-intrusive layer-wise readout architecture that attaches read-only observation heads to selected Transformer layers without feeding their outputs back into the backbone. We consider three variants: a shared LM head across layers, layer-specific heads, and a layer- or step-adaptive head. For full-vocabulary readout, we derive a closed-form overhead approximation governed primarily by V/(12d), with representative estimates ranging from about 30% to 240%, and discuss cost reductions via sparse observation, low-rank factorization, reduced vocabularies, top-k readout, and selective positions. We argue that Bypass Observation can make model computation more observable while remaining only a partial, potentially misleading projection of hidden states. We further distinguish bypass chain-of-thought from conventional chain-of-thought: conventional reasoning tokens enter the autoregressive computation, whereas bypass readouts remain causally external at inference time, although they can still provide training signals in reinforcement learning. Finally, we discuss applications to looped and recurrent-depth Transformers, where iteration-wise readout may expose convergence, oscillation, and potential halting signals. The proposal is conceptual and analytical; systematic empirical validation remains future work.
Read the full abstract
Computation and Language
Understanding the Limits of Agentic ICD Coding
ICD-10-CM codes are alphanumeric codes used in the US to classify diagnoses and injuries for medical billing and epidemiological reporting. Standard ICD-10-CM benchmarks report aggregate metrics that obscure performance on complex coding scenarios. We evaluate neural, workflow, and agentic systems on a rarity-stratified set of MIMIC-IV discharge summaries and identify two orthogonal failure modes. Neural classifiers exhibit a 0.43 micro-F1 gap between rare and common codes. Workflow systems handle rare codes well but score near zero on injury and external cause codes that require multi-step guideline following. A tool-augmented agentic configuration with structured access to official ICD-10-CM reference materials recovers up to 0.34 micro-F1 on this subset. No single system dominates across all conditions.
Read the full abstract
Artificial Intelligence
LLM-Enhanced Multi-Agent Reinforcement Learning for Unified Electric Vehicles-Charging Station-Grid Optimization in Public Charging Systems
In the era of the Internet of Things (IoT), coordinating connected electric vehicle (EV) charging scheduling to balance EV charging satisfaction, station profitability, and smart grid stability presents a complex multi-objective challenge. Existing Multi-Agent Reinforcement Learning (MARL) approaches often struggle with high-dimensional state spaces generated by massive IoT sensing data and conflicting stakeholder interests. This paper proposes a novel LLM-enhanced MARL framework that, for the first time, simultaneously optimizes the Grid, EVs, and Stations within a unified loop. By integrating Large Language Model (LLM), we address two critical bottlenecks: interpretable feature selection and adaptive multi-objective balancing. The LLM analyzes real-time IoT-collected environmental states to extract physically significant features and dynamically assigns weights to conflicting objectives-including profit, user satisfaction, and grid load-using semantic reasoning instead of complex manual tuning. Extensive experiments demonstrate that our framework significantly outperforms state-of-the-art baselines, achieving superior market efficiency while reducing training time by over 70%. This approach offers a scalable, transparent solution for efficient and sustainable IoT-enabled urban charging infrastructure management.
Read the full abstract
Computer Vision
StepPrune: Adaptive Sequential Visual Token Selection across Multimodal Large Language Models
Visual prefixes account for a major portion of the per-layer computation in multimodal large language models (MLLMs), making visual-token pruning a direct approach to accelerating inference. Existing top-K methods typically evaluate tokens independently and apply a uniform budget to all inputs, overlooking both selection-dependent interactions and variations in visual complexity across samples. In contrast, we propose StepPrune, which formulates visual-token pruning as an adaptive sequential decision process. Conditioned on previously selected tokens and textual context, StepPrune progressively constructs the retained subset and automatically determines its size through a learned STOP action. During training, a variance-preserving noise gate provides a differentiable surrogate for the discrete selection process, whereas during inference, unselected tokens are physically removed before language-model prefill. A grouped selection mechanism further extends StepPrune to high-resolution inputs. Experiments across LLaVA-1.5, LLaVA-NeXT, Qwen2.5-VL, and InternVL3 show that StepPrune achieves the best average normalized performance retention across all evaluated pruning rates on LLaVA-1.5, Qwen2.5-VL, and InternVL3, while remaining competitive on the substantially longer AnyRes prefixes of LLaVA-NeXT. On LLaVA-1.5, StepPrune retains 94.6% of the full-prefix normalized performance while pruning 88.9% of the visual tokens. At a mean retained count of 64, StepPrune reduces prefill latency from 59.95 ms to 40.05 ms, corresponding to a 1.50x prefill speed-up.