Information Retrieval
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 3 September
A model that inherits one-line memories may pull one archived source record before acting; a directive in the store can steer that pull: a pointer, a criterion or both. Across sixteen registered studies (179,352 attempts) we measured where the request goes under each form; every result is descriptive, with registered intervals, no mechanism claim. A length-matched criterion exceeded a bare id on six direct-provider models (D) and failed its registered superiority rule on a nine-model OpenRouter panel (E). On generated worlds (K2-K5): the two registered signatures held on Opus 5 and Fable 5.1, Fable 5 followed the same sign, Haiku 4.5 reversed, and Sonnet 5, the GPT-5.6 endpoints and GPT-6 Astra lay near zero (K2). With a defensive adapter at five gains, the 70B rule for a gain-dependent change of the composite - criterion contrast was not met (K3 and K4); under the 8B attenuation rule (0.95 intervals: slope below zero; change beyond the margin), the 8B change of -17.5 [-26.7, -8.1] did not meet it on 36 families (K4) and at registered power on 337 families -16.6 [-19.4, -13.8] did (realised one-sided error at the margin 1.8 to 3.2% per corner of a finite grid, nominal 2.5%, not a uniform-error guarantee; K4's status stands; K5, first ladder), while a second SecAlign++ adapter under the imposed Meta-SecAlign template did not (-11.8 [-14.3, -9.3]; K5, second ladder); no NOT-MET is a statement that the contrast was unchanged; their difference (+4.7 [+2.3, +7.2]) describes two fixed execution paths, licenses no superiority, equivalence or 'significant difference' claim; nothing follows from the statuses differing (K5). Intervals describe family-reweighting stability conditional on the execution, not reproducibility across engine executions; audit replays were neither substituted for nor averaged into outcomes; no missingness gate fired and directional completions changed no status.
Category Theory
The Lie group-Lie algebra correspondence in tangent categories
Classic Lie theory establishes a correspondence between Lie groups, which are internal group objects in the category of smooth manifolds, and Lie algebras. An analogous correspondence also exists for group objects in affine schemes. Both smooth manifolds and affine schemes form tangent categories, which provide a categorical context for differential geometry. Therefore, it is natural to ask whether the Lie correspondence can be constructed entirely from the tangent structure. Building on work of Cockett and Schwarz, we develop an internal Lie group-Lie algebra correspondence in tangent categories. We introduce Lie group objects as group objects in a tangent category which admit a tangent space at the unit. For any group object, we construct a Lie functor and show it is tangentially representable exactly when the group object is a Lie group. Representability then yields an internal Lie algebra whose underlying object is the tangent space at the unit. We prove this Lie algebra is a differential Lie algebra, with a bilinear Lie bracket induced by the adjoint representation, and we extend the construction to a functor from Lie groups to differential Lie algebras. Finally, we recover the usual Lie correspondences in both differential and algebraic geometry.
Read the full abstract
Number Theory
Prime-power Diophantine tuples
A positive $D(n)$-$m$-tuple is a set $A=\{a_1,\ldots,a_m\}$ of distinct positive integers such that $a_i a_j+n$ is a square for every $i\ne j$. In 2005, Dujella and Luca obtained an absolute bound for the cardinality of a $D(p)$- or $D(-p)$-tuple of positive integers, uniformly in the prime $p$. We extend the underlying gap argument to prime powers. An explicit toric elimination certificate, proved by elementary linear algebra, replaces the unsaturated homogeneous elimination step and is valid modulo every prime power. Explicit degree and height bounds for this eliminant yield a uniform gap principle. Combined with the general bound for bounded $|n|$, this shows that every positive reduced $D(\pm p^r)$-tuple (i.e. tuple with elements not divisible by $p$) has less than $2^{121}$ elements, independently of $p$ and $r$. Thus, every positive $D(\pm p)$-tuple has at most $2^{121}$ elements, positive $D(\pm p^2)$-tuples have less than $2^{122}$ elements, and the maximal cardinality of a positive $D(\pm p^r)$-tuple is $O(r)$ uniformly in $p$.
Read the full abstract
Computer Vision
STARS-GS: Structure-Aware Regularized Gaussian Splatting for Large-Scale Aerial Surface Reconstruction
Large-scale 3D surface reconstruction from aerial imagery is fundamental to geospatial mapping and urban modeling. Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated considerable potential for this task. However, existing methods still face three major challenges in large and complex scenes: scene partitioning may split continuous scene elements across independently optimized sub-regions; geometric constraints mainly focus on the attributes of individual Gaussians while overlooking their local organization; and uniform regularization struggles to accommodate heterogeneous geometric structures. To address these issues, we propose STARS-GS, a structure-aware 3DGS framework for large-scale surface reconstruction. First, we introduce a structure-aware scene partitioning strategy that better preserves continuous scene structures during partitioning and reduces cross-region geometric inconsistencies and stitching artifacts through boundary refinement. Second, we develop neighborhood-aware Gaussian organization that extends geometric constraints from individual primitives to their neighborhood organization, encouraging Gaussians to better conform to local surface geometry. Third, we introduce adaptive surface regularization that adjusts the regularization strength according to local geometric characteristics, promoting geometric consistency in structured regions while preserving plausible variations in unstructured regions. Extensive experiments on large-scale aerial photogrammetry benchmarks demonstrate that STARS-GS consistently outperforms the evaluated Gaussian-based methods in surface reconstruction. It increases the average F1-score from 0.640 for the second-best method to 0.698, corresponding to a relative improvement of approximately 9.1%, demonstrating effective improvements in geometric accuracy and surface completeness.
Read the full abstract
Computer Vision
Preserving Knowledge across Space and Time for Continual Video Deepfake Detection
The continuous emergence of high-quality video deepfakes requires detectors that continually adapt to new forgery patterns, yet existing approaches, which are designed for deepfake images, fail to capture video-specific cues. Unlike deepfake images that contain only spatial artifacts, deepfake videos leave distinct evidence along both spatial and temporal axes, necessitating the separate preservation of each modality during sequential model updates. To overcome this limitation, we introduce a continual deepfake video detection framework, Modality-Specific Frequency Distillation (MSFD), that explicitly decomposes video features into spatial, temporal, and spatiotemporal modalities in the frequency domain. This decomposition enables independent preservation of each modality, as different deepfake video types exhibit varying reliance on spatial and temporal cues across tasks. Furthermore, MSFD adopts a cross-modality decorrelation loss that encourages spatiotemporal representations to remain orthogonal to single-modality cues. Extensive experiments show that our framework achieves stronger adaptation and preserves performance more effectively than state-of-the-art methods across diverse continual deepfake video scenarios.
Read the full abstract
Computer Vision
OCR-EDR: Rendering-Aware Diagnosis and Repair for Closed-Loop OCR Improvement
Although document OCR systems perform increasingly well on routine documents, complex formulas, structured text, and long-tail formats remain error-prone. OCR predictions may omit fine-grained content or hallucinate unsupported outputs, while equivalent encodings of the same visible content must be accommodated. Existing OCR evaluation methods mostly report aggregate metrics, offering limited support for analyzing case-level errors and improving OCR performance. We propose OCR-EDR (OCR Error Diagnosis and Repair), a rendering-aware framework that advances from fine-grained diagnosis to iterative repair. Given a source image, an editable OCR prediction, and its rendered image, OCR-EDR first jointly assesses whether the prediction and its rendering are consistent with the source, preserving valid predictions, including rendering-equivalent ones, while diagnosing and localizing genuine errors. It then applies executable edits and may request an updated rendering for iterative reassessment. We construct OCRErrBench from diverse real OCR predictions, covering text and formulas, exact and rendering-equivalent positives, and genuine errors, and develop the DocEDR model to execute the diagnosis–repair loop. On OCRErrBench, DocEDR achieves 94.78% diagnostic accuracy. It repairs 86.23% of erroneous inputs to visual consistency, raises formula Case-F1 by 30.99 percentage points over DOCR-Inspector-7B on DOCRcaseBench, and improves formula CDM by up to 4.62 percentage points on the identified Bad subsets of four OCR systems on UniMER-Test. These results show that OCR-EDR turns fine-grained OCR analysis into verified corrections and performance gains.
Read the full abstract
Number Theory
Odd denominators in the Lonely Runner spectrum for six speeds
For distinct positive integers v_1, ..., v_n let ML(v_1, ..., v_n) be the largest number L such that at some time t every t vi is at distance at least L from the nearest integer; the Lonely Runner Conjecture asserts that ML >= 1/(n+1). Write ML = p/q in lowest terms. Kravitz conjectured that whenever ML < 1/n one has q = np + 1; Fan and Sun found counterexamples for n = 4, conjectured that q = np + k with 1 <= k <= n always holds, and observed that in their data for n = 6 only k = 1 and k = 3 occur. We explain this observation. For six speeds we show that all but finitely many tuples with ML < 1/6 satisfy ML = (P-1)/(6P) for an integer P congruent to 1 or 5 modulo 6; in particular k is 1 or 3 and the denominator q is odd. The proof determines the three infinite two-parameter families of tuples on which such values concentrate, computes ML exactly on each family, and describes exactly where k = 3 occurs. An exhaustive search over the 2 x 10^9 sextuples with speeds at most 110 finds no exception. For five speeds the same method, together with Chen's classification of the tuples attaining ML = 1/5, shows that all but finitely many tuples with ML < 1/5 satisfy Kravitz's original conjecture. The computations are exact and the code is provided.
Read the full abstract
Machine Learning
Beyond Straightness: Non-Crossing Flow Matching via Quantile AlignTree Coupling
The performance of Flow Matching largely depends on the quality of the coupling between the source and target distributions. However, independent coupling often leads to path crossings and local velocity ambiguity, while OT-based couplings typically incur high construction costs. To address this challenge, we propose Quantile AlignTree Flow Matching (QAT-FM), an efficient structured coupling strategy that constructs a hierarchical coupling between a Gaussian prior and the target data distribution via a quantile-aligned tree structure. QAT-FM constructs the coupling in $\mathcal{O}(Nd\log N)$ time and supports per-pair source sampling with $\mathcal{O}(d)$ complexity, enabling scalable training for large-scale high-dimensional generative tasks. Theoretically, we prove that the QAT coupling satisfies marginal consistency, induces non-crossing linear interpolation paths, and consistently improves path separation at intermediate times compared with independent coupling, thereby alleviating local velocity ambiguity. QAT-FM further extends naturally to conditional generation, enabling structured conditional coupling while preserving global Gaussian alignment. Experiments across diverse benchmark datasets demonstrate that QAT-FM achieves competitive generative performance while substantially reducing coupling construction cost.
Read the full abstract
Machine Learning
Guide, Not Bind: Why Defeasible Priors Fail in Augmented Lagrangian Causal Discovery
Differentiable causal discovery methods increasingly encode expert priors as forbidden-edge constraints enforced by an Augmented Lagrangian (ALM) penalty, on the assumption that a data-adaptive relaxation mechanism will discount and eventually override a rule the data consistently contradicts. We show this design, which we call guide, not bind, fails for two independent, precisely characterized reasons, and that directly repairing both restores it only partially. First, sequential penalty-ramping ALM suppresses a wrongly-forbidden true edge before any counterfactual check can detect it: we give three necessary conditions any adaptive relaxation must satisfy to avoid this (Proposition~\ref{prop:conditions}), prove that DADU—the natural relaxation rule this paper introduces as the object of study—violates all three (Corollary~\ref{cor:dadu_failure}), and confirm the failure across 3{,}072 training runs spanning graphs from 4 to 32 nodes, where a single wrong prior suppresses a true edge in 87–97% of trials under DADU. Second, and independent of any fix to the mechanism, we prove in closed form that the standard correlation-matching objective ties a true edge and its reverse to an identical cost of exactly $2r^2$ (Lemma~\ref{lem:tie}), not because the underlying equal-variance model is unidentifiable, but because normalizing to correlation discards exactly the variance information that would make it identifiable; covariance matching instead separates the two directions by a provable margin of at least $w_0^4$ (Lemma~\ref{lem:separation}).
Read the full abstract
Databases
Does SRL Pave the Road to Explainable Reasoning? Lessons Learned from an Implementer's Perspective
The Shape Rules Language (SRL) Working Draft defines how to derive new RDF triples from an RDF graph using inference rules. Each rule matches graph patterns and instantiates triple templates whose output feeds into validation pipelines, SPARQL queries, or further inference. RDF reasoning has traditionally relied on fixed entailment regimes (RDFS, OWL), rule-based ad-hoc languages such as N3, or other implementation-specific solutions without a shared standard. SRL introduces user-defined production rules with a defined grammar, dependency analysis, execution ordering, and termination guarantees. However, no authoritative implementation exists, leaving practitioners with little guidance on how to build a conformant engine or on what problems the language can solve. We implemented two SRL engines and evaluated both on classical RDF reasoning tasks for soundness, completeness, and speed. The first reuses an existing SPARQL query engine and its query parser; the second is a dedicated engine. The SPARQL-based engine reused an existing modular parser for query construction and SPARQL CONSTRUCT for triple production, reducing engine-specific work. The dedicated engine was two to six times faster, the gap widening as rule sets grow. Both engines were validated against the SRL conformance test suite, supplemented by additional use-case-driven tests. A usable SRL engine can be built inexpensively on top of a SPARQL engine, with a moderate speed trade-off that a dedicated implementation recovers. Despite the specification's immaturity, the language already supports practically useful reasoning tasks.
Read the full abstract
Quantum Physics
Convolution absorbing boundaries for explicit-circuit quantum simulation of the wave equation
Explicit quantum circuits for the wave equation, built by Hamiltonian simulation, are restricted to closed domains, in which outgoing waves reflect off the edge of the computational region and return. We lift that restriction with absorbing boundaries of the convolution type - a complex-frequency-shifted perfectly matched layer realised through exponential-kernel memory variables - and make the resulting non-unitary dynamics quantum-implementable by Schrodingerisation. We obtain three results. First, explicit gate-level circuits for the absorbing evolution, obtained by extending a Bell- basis term-evolution circuit to projector-valued operator strings and Trotterising to second order; these run end to end on seventeen to twenty-one qubits and agree with exact references to within a tenth of a percent to a percent. Second, a structural obstruction: the memory form of the absorbing generator carries an irreducibly indefinite Hermitian part, whose largest eigenvalue grows as the square root of the absorption strength divided by the grid spacing and survives any diagonal rescaling of the memory fields. The known recovery threshold for Schrodingerisation then makes the post-selection cost grow exponentially in the simulated time. Third, a remedy: a Lyapunov symmetrizer, precomputed classically, renders the transformed generator dissipative and replaces that time-dependent penalty with a time-independent conditioning factor of several hundred, measured on grids of up to four thousand unknowns and saturating under mesh refinement. The crossover is early: past it the symmetrized recovery is cheaper by four to twenty-six orders of magnitude in post-selection cost, and at the longest horizons tested it is the only recovery that works.
Read the full abstract
Theoretical Economics
Knowledge-Based Mechanisms
We study robust mechanisms when the designer possesses a Bayesian belief over some components of agents' private information but faces ambiguity over others. The designer evaluates mechanisms by their worst-case performance over all joint distributions consistent with her belief over the Bayesian components. The framework encompasses settings such as multidimensional delegation in which a principal knows the distribution of the state but not the agent's preferences (e.g., his tradeoffs across dimensions), screening in which a seller only has misspecified estimates of buyer preferences, and auction and voting design when agents' beliefs about each other are ambiguous to the designer. We provide conditions under which a knowledge-based mechanism—one that conditions only on the Bayesian components but not the ambiguous ones—is robustly optimal. Our results unify earlier work across distinct economic environments and uncover new applications.
Read the full abstract
Artificial Intelligence
Do GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI Agents
Graphical user interface (GUI) agents are increasingly used to execute natural-language instructions on user interfaces, yet real users may issue infeasible instructions due to benign mistakes. A reliable agent should not only know how to act, but also when not to act. In this work, we introduce CONFLICTGUI, a benchmark covering instruction-internal conflicts and instruction-GUI context conflicts to study conflict-aware termination. Our evaluation reveals severe execution-biased overcompliance: agents that perform well on feasible tasks often continue to execute blindly under conflicting instructions. To mitigate this behavior, we propose CONFLICTGUARD, an inference-time framework that aligns an agent's feasibility awareness with its action generation. CONFLICTGUARD contains two coupled components: a feasibility verification protocol that guides the agent to assess instruction logic and GUI-side evidence before acting, and a conditional action modulation mechanism that steers agents from over-compliant execution into termination-oriented behavior. Experiments across five widely-used agents demonstrate that CONFLICTGUARD improves average conflict task success rate significantly, while preserving normal GUI-task performance. These results validate that a lightweight inference-time intervention can substantially boost GUI Agent's competence to identify inappropriate execution scenarios and refrain from unnecessary actions.
Read the full abstract
Computational Engineering
Markovian Shock-Source Tracing and Multidimensional Asset Roles in Exchange Rates, Gold Futures, and Bitcoin
This study examines cross-asset connectedness in an international financial network of major exchange rates, gold futures, and Bitcoin. Moving beyond the conventional net transmitter–receiver classification, we characterize asset roles through three complementary dimensions: direct spillover transmission, stationary source-tracing dynamics, and multistep upstream connectivity. Return spillovers are estimated using VAR generalized forecast-error variance decomposition (VAR-GFEVD). The positive pairwise net-spillover structure is then mapped into a row-stochastic Markov kernel whose transitions trace dominant net-spillover sources in the reverse direction of the original transmission edges. A stationary departure flux describes long-run movement in this source-tracing chain, while Viral Centrality is evaluated by the deterministic probability-propagation algorithm of Fink et al. to approximate multistep upstream reach. Empirically, gold futures emerge as the dominant direct net transmitter and the leading stationary source-tracing node, but do not have the largest Viral Centrality. Several exchange-rate nodes classified as direct net receivers have relatively high Viral Centrality, indicating broad conditional access to upstream source nodes. Bitcoin occupies an intermediate role. An auxiliary specification including the U.S. Dollar Index yields qualitatively similar role differentiation. The results show that direct connectedness, stationary source tracing, and approximate multistep upstream reach cannot be inferred from NET spillovers alone. All diffusion interpretations are descriptive and conditional on the estimated network, row normalization, and restart closure.
Read the full abstract
Machine Learning
It's the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories
Reasoning traces of large language models are widely read as containing "breakthrough" moments and early-legible fates. Both readings rest on measurements missing a counterfactual control at the level of the claim; we supply both controls. First, a restart-controlled truncation probe separates when a solution fits the continuation budget from when a prefix carries value that fresh computation cannot buy, comparing per-anchor continuation solve rates against from-scratch restart curves at matched total generated-token budget. Applied to 178 problem-model cells (89 MATH problems x two small open models, an outcome-blind but difficulty-targeted cohort), exactly 1 of 178 cells survives as prefix-limited; restart dose-response separates a compute-starved model from a capability-limited one; and wherever the matched budget lies inside the restart grid, continuing the model's own prefix beats restarting (9 of 9) – predominantly compute compression rather than expanded reachability. Second, a pre-registered, difficulty-controlled test finds no detectable outcome information in early-window internal signals beyond a problem-difficulty baseline, and two generation-free analyses of public corpora show why this control is needed: a trace-blind difficulty proxy reaches AUROC 0.873 on 192K DeepSeek-R1 generations – inside the published probe range – and a closely matched reconstruction of the closest published early-window positive recovers a comparable pooled result (0.849) while within problem it is statistically indistinguishable from chance at all ten anchors (0.496 at t=4); a post-hoc within-targeted probe finds only a small average residual, concentrated in three low-failure problems. High pooled probe AUROCs cannot by themselves establish within-attempt information; a question-only baseline or within-problem evaluation is required.
Read the full abstract
High Energy Astrophysical Phenomena
The Jet Properties and Accretion Regime for the Blazar Sequence
The puzzling bimodality displayed by blazars on the broad-band spectral indices ($\alpha_{\rm ro}-\alpha_{\rm ox}$) plane is still an open question. To investigate its physical origin, we compiled a sample comprising 136 flat-spectrum radio quasars (FSRQs), 64 low-synchrotron-peaked BL Lac objects (LBLs), 39 intermediate-synchrotron-peaked BL Lac objects (IBLs), and 105 high-synchrotron-peaked BL Lac objects (HBLs). Our results show that (FSRQs+LBLs)$\rightarrow$IBLs$\rightarrow$HBLs follow a \reflectbox{$\angle$}-shaped evolutionary sequence on the $\alpha_{\rm ro}-\alpha_{\rm ox}$ plane, with the average 5\,GHz radio Doppler factor, jet power, and accretion Eddington ratio decreasing along this \reflectbox{$\angle$}-shaped track. Additionally, the distributions of $\alpha_{\rm ox}$ exhibit only slight differences among blazar subclasses, whereas those of $\alpha_{\rm ro}$ and $\alpha_{\rm rx}$ differ significantly. Our results can be naturally explained if the distribution of blazars on the $\alpha_{\rm ro}-\alpha_{\rm ox}$ plane is regulated by the accretion rates and the Doppler beaming effect across different bands ($\delta_i^{q_i}/\delta_j^{q_j}$). However, $\delta_i/\delta_j$ plays a pivotal role in the distribution, while jet shape ($q=2(3)+\alpha_\nu$) and spectral indices ($\alpha_\nu$) play a minor role. Our results suggest that FSRQs+LBLs are strong-jet sources coupled with radiatively efficient accretion, and their jets are accelerated to a larger distance from the core and still maintain relativistic speeds down to the radio region; HBLs are weak-jet sources coupled with radiatively inefficient accretion, and their jets may begin to decelerate in the optical region and exhibit sub-relativistic speeds as they propagate to the radio region. IBLs are regarded as transitional objects between FSRQs+LBLs and HBLs.
Read the full abstract
High Energy Physics (Phenomenology)
CGC-py: A Monte Carlo Event Generator for Gluon Saturation Physics
We develop CGC-py, a Monte Carlo event generator for deep-inelastic scattering. It couples the full Color Glass Condensate (CGC) cross section for $\gamma^*p(A)\to q\bar q+X$ to a Parton-Branching transverse-momentum-dependent backward initial-state shower, while Pythia~8 handles final-state radiation and hadronization. CGC-py retains the complete target-elastic and target-inelastic contributions without taking the back-to-back correlation limit, allowing single- and di-hadron observables to be generated consistently from the same event sample. We validate the generator through an analytic closure test of the single-inclusive quark spectrum and a comparison of charged-hadron spectra in $ep$ collisions with H1 data, finding excellent agreement. The predicted nuclear modification factor $R_{e\mathrm{Au}}^h$ shows the expected saturation pattern: suppression at low $p_T^*$ followed by a rise toward unity at higher $p_T^*$. A comparison with a Pythia~6 baseline, together with an $x_g$-rescaling study, indicates that small-$x$ CGC evolution and collinear DGLAP dynamics contribute comparably to the growth of the dihadron away-side width with energy. Genuine saturation-driven broadening emerges only at the highest energies considered. Within CGC-py, $e\mathrm{Au}$ collisions exhibit an enhanced away-side width and a suppressed back-to-back yield relative to $ep$ collisions. These nuclear effects remain modest over EIC kinematics, motivating measurements at the most forward accessible kinematics and the use of complementary observables to maximize sensitivity to gluon saturation.
Read the full abstract
Quantum Gases
Universal scaling of fluctuations and correlations across the superfluid transition
The critical behavior of a system near a continuous phase transition is commonly characterized by power laws and a set of universal critical exponents. Yet universality at phase transitions has even more fundamental manifestations. It dictates that all observables follow universal scaling functions, which may extend beyond simple power laws. Furthermore, microscopic details are captured entirely by only two system-specific scale factors, that rescale any measured quantity onto its universal form. In this work, we report the experimental validation of this prediction in ultracold lattice Bose gases undergoing the superfluid transition. We extract the scale factors by collapsing the order-parameter cumulants measured at varying entropies. We show that these factors specify scaling functions governing both the order-parameter distribution and two-point correlations. By measuring these scaling functions, we also determine three critical exponents $\beta$, $\gamma$ and $\nu$. Our results provide an experimental demonstration that universal scaling is simultaneously obeyed by multiple observables in a quantum many-body system.
Read the full abstract
Computation and Language
Decoupled Analysis-Judging: An Automated Creativity Evaluator Using LLMs in Complex Multi-step Creativity Tasks
Automated evaluation of creativity tasks remains challenging for LLM-as-a-Judge, as LLM is susceptible to biases such as verbosity bias and leniency bias. Such limitations are particularly evident in Contextually-Grounded and Procedurally-Structured Tasks (CGPST), a complex multi-step creativity task where inter-step dependencies, highly subjectivity, and wide scoring ranges lead to more unstable and biased judgments. Existing approaches either rely on task-specific training or directly apply LLM-as-a-Judge, both of which struggle to ensure reliable evaluation under such complexity. To bridge these gaps, we propose CreaEval, an automated creativity evaluator for CGPST that decouples typical LLM-as-a-Judge into analysis and judging. Correspondingly, CreaEval involves two critical phases: Memory-augmented Analysis, a SoT-LLM converts multi-step responses into structured evaluation evidence, incorporating cross-step memory; and Evidence-based Judging, a Judge-LLM uses the extracted evidence for judging without accessing raw responses. Comprehensive experiments show that CreaEval achieves an average performance improvement of 22.74% over the second-best baselines across CGPST and two classic simple creativity tasks, demonstrating its generalizability. The code is available at https://github.com/Jaong/CreaEval.
Read the full abstract
Astrophysics of Galaxies
Constraints on mu-variations from the methanol CH3OH thermal lines in the Galaxy at galactocentric distances of 1.5 < D_GC < 13 kpc
In this study, constraints are set on hypothetical variations of the fundamental physical constant mu = m_e/m_p - the electron-to-proton mass ratio - based on thermal emission lines of E-methanol (CH3OH) in the 216-220 GHz frequency range. Molecular spectra toward seven massive clouds located across a wide range of galactocentric distances (1.5 < D_GC < 13 kpc) have been analyzed. The obtained weighted mean value <Delta mu/mu> = (0.5+/-0.5)*10^(-7) does not indicate statistically significant changes in mu at the 1 sigma confidence level of 5*10^(-8), which is in good agreement with previously established upper limits in a number of molecular clouds in the Galactic disk.
Read the full abstract
Computation and Language
Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
Large language models achieve superior performance on tasks that require extended reasoning, but long chains of thought make the KV cache a severe memory bottleneck. Existing KV cache compression methods share one paradigm: score each cached token by some estimate of how much it will matter later, and keep the top-scoring ones. We show that the selection signal contributes almost nothing. Random Attention keeps the prompt and evicts uniformly at random within each attention head, computing no score at all; across four models and six reasoning tasks, it matches the strongest baseline in task performance while delivering 32-43% higher throughput than that method when deployed with vLLM. Controlled experiments explain this by showing that 1) the prompt is the fragile part of the cache, and most of the gap between selectors is just whether their selection signal happened to keep it; 2) the reasoning trace protects itself against eviction with redundancy at two levels, in the text (the model restates what it still needs as it works) and across attention heads (each keeps its own copy of the trace), so once the prompt is safe, a random draw retains enough copies of what the model still needs, and no score is required to pick them. Our code is publicly available at https://github.com/SalesforceAIResearch/Random-Attention.
Read the full abstract
Computer Vision
When Do Frozen VLMs Respond to Image-Free Object-Token Edits? An Answer-Key-Free Protocol and What It Reveals
Answering what-if queries about a scene with a VLM usually means injecting the assumption as text or repainting the scene with a generative model. We instead move the edit to the representation level, before the model input. The image is abstracted into a set of object-level tokens, and the original image never enters the VLM. This design rests on an open question: when do frozen VLMs actually respond to such token edits? We introduce an answer-key-free protocol: no post-edit answer is annotated. It scores edits whose answers are logically determined, and audits itself by reversing each scoreable choice. The protocol reveals three structures. The response is not free: explicit edit teaching, not ordinary VQA training, produces it in dense scenes and multiplies it in sparse ones, on all three operations. Once on, it is governed by token cleanliness and density, with deployable detector+segmenter tokens competitive with the oracle and outperforming it on VRSBench. And reading is a separable axis: the image-free token route preserves 92-96% of a matched patch-token baseline's free-text VQA, and the answers measurably depend on the tokens. The response, cleanliness, and reading structures are sign-preserved across two remote-sensing datasets (iSAID, VRSBench) and three frozen LM backbones. We release the probe generator, records, judge logs, and code.
Read the full abstract
Combinatorics
A New Sufficient Condition for Oriented Graphs Determined by Their Generalized Skew Spectra
Characterizing graphs uniquely determined by their spectra (DS) is a core open problem in spectral graph theory. While this problem has been extensively investigated for simple undirected graphs, it remains relatively underexplored for oriented graphs. For a simple undirected graph $G$ equipped with an orientation $\sigma$, the corresponding oriented graph $\Sigma=(G,\sigma)$ is the digraph obtained by orienting each edge of $G$ according to $\sigma$. An oriented graph $\Sigma$ is said to be determined by its generalized skew spectrum (DGSS) if every oriented graph sharing the same generalized skew spectrum is isomorphic to $\Sigma$. This paper develops a new sufficient criterion for recognizing DGSS controllable oriented graphs, which applies to a much broader family of graphs than previously known results. Let $S$ be the skew-adjacency matrix of $\Sigma$, $W(\Sigma)=[e,Se,\ldots,S^{n-1}e]$, and $d_n$ the last invariant factor of $W(\Sigma)$. For each odd prime $p$, we define the polynomial $\Phi_p(\Sigma;x)=\gcd(\chi(S;x),\chi(S+J;x))$ over the finite field $\mathbb{F}_p$, which is invariant under generalized skew cospectrality. By analyzing the square-free part of $\Phi_p(\Sigma;x)$ and the associated $p$-main polynomial, we establish a DGSS sufficient condition under the square-free assumption on $d_n$. The proposed criterion allows higher $p$-nullity and recovers the square-free determinant criterion of Qiu, Wang and Wang~(2019) as a special case. We further provide illustrative examples to verify the wider applicability of our new condition and to highlight the role of the compatibility constraints on the irreducible factors of $\Phi_p(\Sigma;x)$.
Read the full abstract
Machine Learning
TraveL: Transformer-based Multi-view Path Distributional Representation Learning
Path representation learning (PRL) for road networks has received increasing research attention, due to various path-related applications. Existing works on PRL typically exploit the co-occurrence relationship among road segments and paths to learn a vector as the path representation, without exploring the varied traveler behaviors and the regional correlation on the path. In this work, we propose to learn distributional representations, which provide valuable information for use in path-related applications, by capturing the varied traveler behaviors as well as the various dependencies within regions of road segments. We propose a novel Transformer-based Multi-view Distributional Representation Learning (TraveL) framework to encode a path along with a travel starting time to a distributional representation, which can be used to decode possible samples of on-path traveler behavior. Moreover, by analyzing the regional correlation which reveals various road segment relationships, we propose a regional attention to encode these correlations in a path. Also, we explore the idea of Kolmogorov-Smirnov (K-S) test to compare the sampled traveler behavior against the collected ground truth to facilitate training. Experimental results show that the proposed TraveL model outperforms the state-of-the-art methods on both synthetic and real-world datasets, by 14.7% in Mean K-S distance for travel time distribution estimation, 16.7% in Mean Absolute Error (MAE) for path similarity prediction, and 3.97% in MAE for destination prediction.
Read the full abstract
Computation and Language
Lngram v2: Latent N-Gram Memory with Interpretable Discrete Representations
Transformers lack a native lookup mechanism, requiring repeated dense computation to recognize and reuse local static patterns. Lngram v1 introduces tokenizer-independent conditional memory through discrete latent n-gram addressing, but its memory capacity is coupled with the backbone width, limiting scalability due to high parameter and activation costs. We propose Lngram v2, which decouples the number of routes, memory dimension, and backbone width, and introduces a context-aware grouped-query attention readout to scale memory capacity independently. A zero-value Sink and counterfactual surrogate gradients further improve readout selectivity and routing trainability while preserving hard discrete addressing. Experiments across vision–language models (VLMs) of different scales show consistent improvements, including successful scaling to a 30B-parameter model. Compared with Lngram v1, Lngram v2 substantially reduces both total and activated memory parameters while maintaining or improving language modeling performance. Further analysis shows that its discrete IDs preserve substantial semantic structure of continuous hidden states, enabling semantic recovery from IDs alone and stable ID–semantic associations across datasets. These results establish Lngram v2 as an efficient and scalable latent conditional memory mechanism whose discrete addresses also provide a structured interface for analyzing internal model representations.
Read the full abstract
Multiagent Systems
The Civilization Framework: Sovereign-Anchored Communication Between Personal Multi-Agent Systems
Humans are the transport layer between AI systems, losing context at every hop. We present the Civilization Framework, whose addressable party is the civilization, not the agent (one human sovereign, a persistent ledger, and interchangeable agents), and the Embassy Protocol, a carrier-agnostic overlay: messages arrive asynchronously at a resident ledger endpoint, any online agent of the receiver handles them, and commitment state on both ledgers, not delivery, is ground truth. Authority derives from memory: an agent's power to act for its civilization is capped by the memory it can access and externalized through signed credentials, separate from civilization-level reputation. We identify the temporal-weight effect, a hazard in AI-to-AI communication where what arrives first acquires unearned authority, and test it in one frontier model in a preregistered 1,908-trial experiment. With verification removed, an incorrect upstream claim arriving first captures 54.2% of answers (4.2% under full verification), while the same claim arriving after the receiver has sealed its own answer captures 31.6% (the two prompt shells are not length-matched, so part of that gap may reflect shell form; see Section 7), and both registered question-set specifications agree on these two verdicts (the exclusion specification is preregistered as under-powered). Two secondary results, the mitigation from instruction-level provenance labeling and sealed-answer accuracy equivalence, are specification-dependent, holding only under the all-questions specification. Because a registered check of tool use failed its call-budget condition, the registration classifies the round as inconclusive and every result above, primary and secondary, is reported as exploratory; a replication with harness-enforced budgets is planned. The framework's intra-civilization layer has a working implementation.
Read the full abstract
Differential Geometry
Catenoid-sharp index-topology estimates for minimal hypersurfaces
We show that for an embedded two-sided minimal hypersurface in $\mathbb{R}^N$, there is a lower bound for the index in terms of the first Betti number and the number of ends, using ideas from the recent work of Chodosh–Gianocca. This estimate is sharp for the higher-dimensional catenoid. We also obtain a $\operatorname{Spin}(7)$ analogue of Theorem 10.1 of Chodosh–Gianocca: a complete two-sided minimal immersion $M^7\to\mathbb{R}^8$ of index one and finite total curvature is a higher-dimensional catenoid.
Read the full abstract
Artificial Intelligence
DuplexSpeechBench-IFEval: Evaluating Implicit Instruction Following in Full-Duplex Voice Agents
Full-duplex voice agents must continuously decide when to speak, listen, backchannel, interrupt, overlap, and yield the conversational floor. Existing benchmarks evaluate these behaviors through explicit turn-management instructions, whereas voice agents are often configured through roles or personas from which appropriate conversational behavior must be inferred. We introduce DuplexSpeechBench-IFEval (DSB-IFEval), a benchmark for evaluating implicit instruction following in real-time spoken interaction. DSB-IFEval comprises 1,038 test cases derived from 240 controlled conversations spanning eight behaviorally contrastive assistant roles and five conditioning protocols: default behavior, explicit behavioral instructions, persona-implied behavior, combined persona-rule conditioning, and instruction conflict. We measure real-time floor management using the deterministic Instruction Adherence Score (IAS) and persona-consistent response content using the LLM-judged Persona Adherence Score (PAS). Across eleven real-time speech models, we find that executing an explicit floor-management policy does not reliably imply the ability to infer the same policy from a persona. Moreover, even frontier models such as GPT-Live-1 and Gemini-3.8-Live adapt their dialogue language to the assigned persona without consistently translating that persona into the appropriate full-duplex floor-management behavior. Finally, models that successfully resolve benign instruction conflicts often fail when safety-relevant role behavior should override an explicit directive. These results show that inferring role-implied behavior, executing it in real time, and resolving instruction conflicts remain distinct challenges for full-duplex voice agents.
Read the full abstract
Machine Learning
Inferred Generative-Process Diversity Predicts Correlated Failure Across Language Models
Diversity is a widely observed factor in the resilient function of collective systems, yet the type of diversity that matters depends on the properties and failure modes of the system. This distinction is important for systems composed of multiple language models. Different models may be treated as independent components even when their behaviour and failures remain strongly correlated. Assessments of language-model populations using semantic similarity demonstrate limited semantic diversity, but this captures only differences in the meaning of observed outputs. We argue that a more fundamental notion of model diversity is generative-process diversity, the differences between processes capable of generating the observed outputs. Drawing from Algorithmic Information Theory, we use Normalised Compression Distance between raw model outputs, residualised against a permutation control, as a measure of inferred generative-process diversity. Across 38 language models, this measure identifies population structure missed by semantic similarity and predicts cross-task variation in chance-corrected correlated failure among model pairs across ten disjoint benchmark families, beyond semantic similarity and model-pair capability. The cross-benchmark partial rank association is $-0.216$ with a 95% interval of $[-0.309,-0.122]$, and the estimate is negative on all ten benchmarks. These results indicate that increased generative-process diversity is associated with reduced correlated failure in model pairs that is not attributable to semantic similarity or capability. Inferred generative-process diversity offers a novel and practical approach for investigating diversity of multi-model systems in safety-relevant contexts.
Read the full abstract
Quantum Physics
Tunable topological enhancement of covariant quantum Fisher information via non-Bloch skin effect in non-Hermitian SSH lattices
The covariant quantum Fisher information (CQFI) has recently been established as the ultimate precision benchmark for pseudo-Hermitian sensors [Phys. Rev. Lett. 136, 080802 (2026)], yet existing analyses are limited to single-mode systems. Here we extend the CQFI formalism to multi-mode non-Hermitian Su-Schrieffer-Heeger lattices and reveal tunable topological enhancement enabled by the non-Hermitian skin effect (NHSE). Under open boundary conditions, the NHSE deforms the conventional Brillouin zone into a generalized Brillouin zone of radius r = exp(ąppa}), where ąppa} denotes the non-Bloch decay rate. While the total CQFI scales linearly with system size N, its prefactor depends critically on ąppa}, yielding an enhancement factor E(N) = F_OBC / F_PBC that exceeds 30 for N ~ 30, substantially outperforming periodic-boundary sensors. The enhancement is robust against moderate local disorder and supports multi-parameter estimation, with the joint Cramér-Rao bound reduced by up to 15 orders of magnitude. These findings establish a spatial-domain mechanism for quantum metrology that complements time-domain strategies and is experimentally accessible using topoelectrical circuits, photonic lattices, and superconducting circuits with current technology.