Computation and Language
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Wednesday 2 September
Diffusion language models (DLMs) have emerged as a promising alternative to the auto-regressive paradigm. With bidirectional attention and any-order generation, DLMs naturally fit infilling tasks, which require generating a middle span conditioned on both the prefix and the suffix. However, infilling is sensitive to the length of the span, while DLMs require the length to be fixed before generation. Although prior studies extend DLMs to dynamic lengths, they still suffer from two limitations. (i) Sensitivity to initial length. These methods require a preset length to initialize the search and are highly sensitive to this initial length, often yielding suboptimal results. (ii) Inference inefficiency. They either insert length-changing operations during generation or repeatedly search for an appropriate length using multi-step denoising confidence, both of which introduce substantial extra forward passes and computational cost. Therefore, we propose PILL (Probing-based InfiLling with preset-Length-free decoding), an efficient infilling method for DLMs that requires no preset initial length and adds far fewer extra forward passes than baselines, substantially reducing inference time. Experiments show that, across five DLMs spanning different families, architectures, and training recipes on eight infilling benchmarks, PILL improves over the strongest baseline by +4.8 average pass rate on code and +6.0 BLEU-2 on text, while running 1.82x faster than that baseline. The code is available at https://github.com/Hsu1023/PILL.
Machine Learning
A Unified Rate-Distortion Perspective on Vector, Product, and Scalar Quantization
Discrete visual tokenization, predominantly driven by vector, scalar, and product quantization, lacks a unified conceptual framework for understanding quantization tradeoffs. In this paper, we propose a unified rate–distortion perspective on modern discrete visual tokenization. By viewing quantization as lossy compression, we characterize the nominal fixed-length coding rate through token count and codebook size, and quantization error as the distortion. Within this framework, we resolve three central questions. First, we theoretically and empirically show that minimizing distortion, rather than maximizing codebook utilization, is the primary intrinsic objective for reconstruction fidelity, with a direct connection to the STE-induced gradient discrepancy. Second, we establish two critical fairness conditions for intrinsic quantization comparison: controlling latent feature statistics and enforcing identical coding rates. Third, under these conditions, we recover the VQ–PQ–SQ distortion hierarchy in modern visual tokenization and show empirically that modern VQ methods achieve the lowest distortion. This work provides a foundational rate–distortion reframing of modern discrete visual tokenization, resolves ambiguities in quantizer evaluation, and provides a controlled framework for isolating intrinsic quantization effectiveness under fixed-rate constraints.
Read the full abstract
Databases
Git4Data: Database-Native Version Control for AI Agents
Large Language Model (LLM) agents increasingly explore many candidate states of relational data in parallel, each of which should remain isolated, reproducible, and auditable, preferably through the same SQL interface used for ordinary data work. Existing tools support this requirement only partially: source-code version control does not scale to large datasets, whereas relational databases manage large data efficiently but rarely expose native branching, comparison, and merging. We present Git4Data, a database-native version-control layer for agentic workflows. Git4Data treats a database as a repository and a table as a versioned object, exposing Git-style operations (snapshot/tag, branch, diff, and merge with explicit conflict-resolution policies) through SQL extensions. Implemented in MatrixOne, a cloud-native relational database, Git4Data leverages immutable object storage and MVCC to make the cost of these operations proportional to the size of the change rather than the size of the data. On the BranchBench agentic branching workloads, Git4Data outperforms DoltDB by up to an order of magnitude. Overall, we believe this work sheds light on how relational databases can better support AI agents through efficient versioning.
Read the full abstract
Algebraic Geometry
Contact fundamental forms and adjoint varieties
We introduce contact symbol systems, a noncommutative analogue of symbol systems for projective fundamental forms, by replacing the polynomial algebra on a vector space by the graded dual of the universal enveloping algebra of a Heisenberg algebra. For a complex projective submanifold equipped with a contact structure, we define contact fundamental forms and prove that, at a general point, they form a contact symbol system, which gives a contact version of the classical result due to E. Cartan. Conversely, we prove that every contact symbol system can be realized as the contact fundamental forms of a projective variety with a dense open Heisenberg orbit, called the Heisenberg-symmetric variety associated to the contact symbol system. We show that the closure of a projectivized nilpotent orbit in a simple Lie algebra is Heisenberg-symmetric if and only if it is the adjoint variety, namely, the projectivization of the minimal nilpotent orbit. For adjoint varieties of non-symplectic simple Lie algebras, we prove the contact analogue of the Landsberg–Manivel strict prolongation property by using Yamaguchi's prolongation theory.
Read the full abstract
Differential Geometry
Conformal Ricci Curvature and Spectral Estimates
We derive optimized upper bounds for the first Dirichlet eigenvalue of a bounded domain under lower bounds for the conformal Ricci tensor of Shen and Ye. The method also yields sharp estimates for the bottom spectrum on complete noncompact manifolds, a four-dimensional application involving $Q$-curvature, and local and global spectral-Ricci extensions of Cheng's estimate. As a consequence, we obtain a sharp comparison between the spectra of the Laplacian and the conformal Laplacian in terms of the smallest Schouten eigenvalue.
Read the full abstract
Commutative Algebra
On the vertex connectivity of weakly zero-divisor graph of commutative rings
The weakly zero-divisor graph $W\Gamma(R)$ of a commutative ring $R$ is the simple undirected graph whose vertices are nonzero zero-divisors of $R$, and two distinct vertices $x$, $y$ are adjacent if and only if there exists $w\in {\rm ann}(x)$ and $ z\in {\rm ann}(y)$ such that $wz =0$. In this paper, first we prove that the vertex connectivity of $W\Gamma(R)$ is equal to its minimum degree, where $R$ is either an Artinian ring or reduced ring. For any finite ring $R$, we obtain the vertex connectivity of $W\Gamma(R)$. Moreover, this paper characterizes all the vertices that attain the minimum degree of $W\Gamma(R)$.
Read the full abstract
General Relativity and Quantum Cosmology
Cosmological implications for hairy black holes via spontaneous symmetry breaking: Are Hairy Black Holes Primordial?
We investigate whether hairy black holes generated through spontaneous symmetry breaking in Einstein-Scalar-Gauss-Bonnet (ESGB) theory, involving a complex scalar field with a global $U(1)$ symmetry, can be compatible with cosmological evolution. To this end, we introduce the ESGB theory with a scalar self-interaction that becomes relevant on cosmological scales while remaining negligible near the black hole. Owing to the time dependence of the GB term on cosmological scales, the scalar field dynamics in the evolving FLRW background differ qualitatively from those in the nearly static black hole background. In particular, for scalar-GB couplings compatible with hairy black hole formation, the effective potential supports a symmetry-broken vacuum throughout inflation. However, after inflation, a decelerated expansion changes the sign of the GB term, temporarily making the effective potential unbounded from below. As the GB contribution subsequently decreases, the scalar self-interaction eventually dominates and restores the symmetry. Within this schematic framework, we derive stringent constraints on the coupling strengths, the cutoff scale, and the black hole mass, which primarily arise for avoiding efficient tachyonic amplification of the scalar field perturbations during the unbounded phase. For cutoff scales compatible with both cosmological evolution and scalar hair formation, we find that only ultralight black holes with masses of the order of a few grams can develop scalar hair, identifying them as hairy primordial black holes.
Read the full abstract
Machine Learning
Federated LoRA Adaptation of BiomedCLIP Across Four International Chest X-Ray Cohorts
Federated learning (FL) lets institutions train a shared model without exchanging data, and Low-Rank Adaptation (LoRA) makes this practical at scale by communicating only compact low-rank updates. Biomedical imaging is a compelling setting for this combination: patient data are archived behind privacy regulations, and institutions differ widely in scanners, protocols, and compute. Such heterogeneity raises the question of how federated LoRA updates should be aggregated, increasingly pressing as multimodal vision-language models become central to medical image analysis. We benchmark federated Parameter-efficient fine-tuning (PEFT) of BiomedCLIP for chest radiograph classification across four public cohorts on three continents (USA, Vietnam, Spain). Federated LoRA adaptation improves shared-class AUC on all four cohorts over the unadapted BiomedCLIP backbone (mean 0.687 to 0.802), showing that the gains come from federated adaptation rather than from the pretrained model's zero-shot ability. Relative to isolated single-cohort training, federation improves the weaker cohorts while largely preserving the strongest and approaches a centralized reference (0.812) that pools all data. The singular value decomposition (SVD)-based product-space aggregation introduced by FlexLoRA is essential to this gain (naive factor averaging drops mean AUC by 0.097), whereas a drift-correcting optimizer (FedProx) shows no benefit over FedAvg in our single-seed runs, consistent with LoRA's low-rank updates already limiting client drift. Biomedical vision-language models can thus be adapted collaboratively across heterogeneous, geographically distributed institutions without centralizing data.
Read the full abstract
Mesoscale and Nanoscale Physics
Persistence and emergence of quantum defects through pressure-induced phase changes
Extreme pressures can transform materials and their properties, but probing these in-situ is made challenging by the small sample volumes and access requirements demanded by diamond anvil cells. Quantum defects offer a route to local measurements under such conditions, yet their sensing performance can be dictated by pressure-induced changes in their own host material. On the other hand, pressure may also be harnessed as a tool to engineer and stabilize new quantum defects with emergent functionalities. Here, we demonstrate both aspects within a unified platform based on optically active spin-pair defects in hexagonal boron nitride (hBN). As robust quantum sensors under pressure, these spin-1/2 systems retain pressure-independent spin resonances up to 20 GPa while maintaining or even enhancing their optical emission, in stark contrast to the spin-1 boron-vacancy centre in the same material. Simultaneously, we show that compression acts as a means of quantum defect engineering: the starting hBN undergoes an irreversible transformation into wurtzite boron nitride (wBN), during which the defect landscape is reconfigured. Spin-pair sensors are seen to persist across this structural transition, however, depending on the starting material we also observe new, highly fluorescent defects in the wBN phase. These results establish spin-pair defects in boron nitride as pressure-resilient quantum sensors while highlighting high pressure itself as a versatile pathway for creating and tuning quantum emitters.
Read the full abstract
Optics
Laser-induced phase shift of swift electrons
We revisit the calculation of the phase shift experienced by swift electrons on passing through the electromagnetic field of a laser. Such phase shifts are now utilized in the form of `laser phase plates' in transmission electron microscopes (TEMs), for example. We calculate the phase shift using three different methods, namely, perturbation theory applied to the Dirac equation, the Volkov solution to the Klein-Gordon equation, and the relativistic Hamilton-Jacobi equation. We find that all three methods are in agreement, and that the calculated phase shift is independent of the relative orientation of the electron and laser beams. The agreement between the quantum and classical theories is explained. The duality between the phase shifts experienced by electrons, photons and neutrons is described. Our Lorentz invariant result for the phase shift differs from certain results published in the literature.
Read the full abstract
Quantum Physics
An Entanglement-Assisted Stabilizer Framework for Distributed Sensing of Local Phases
Distributed quantum sensing requires spatially separated probes to acquire local parameters while maintaining compatibility with network-level quantum information processing. We develop an entanglement-assisted stabilizer framework based on the extended structure of entanglement-assisted quantum error-correcting (EAQEC) codes, in which the remote halves of pre-shared ebits are used directly as local phase probes while the joint state simultaneously carries an encoded logical subsystem. Each remote probe acquires a local Z-axis phase and subsequently returns through an X-type noise channel. Within the extended EAQEC stabilizer structure, the stabilizer containing $X_{B_j}$ provides phase-dependent measurement statistics, whereas its partner containing $Z_{B_j}$ records the corresponding return-error syndrome. A graph-code formulation is introduced to make this structure explicit, together with an illustrative [[5,1,3;2]] construction. We further show that, conditioned on the joint sensing-and-syndrome measurement record, the post-sensing state differs from the original encoded state only by a known Pauli transformation, so that the logical information remains available for subsequent encoded operations. For the local-phase model considered here, the stabilizer readout attains the available quantum Fisher information, while the finite-shot estimator approaches the corresponding $1/\sqrt{M}$ scaling as the number of repetitions increases. The framework therefore provides a common entanglement-assisted stabilizer structure for distributed local-phase sensing, restricted return-error identification, and post-sensing logical-state retention, without relying on an intrinsic metrological enhancement from EAQEC itself.
Read the full abstract
Computer Vision
Evidence-Guided Detection, Localization and Explanation for Text-Centric Image Forensics
The rapid progress of AIGC has made text-centric image manipulation increasingly accessible, creating new forensic challenges that require not only authenticity detection but also spatial grounding and evidence-based explanation. This paper presents our solution to the GenText-Forensics Challenge at ACM Multimedia 2026. We propose an evidence-guided detector-localizer-reasoner system, where an image-level detector provides a global authenticity prior, a dedicated localizer extracts tampered regions as spatial grounding evidence, and an MLLM-based reasoner generates structured forensic reports grounded in this expert forensic evidence. These modules are connected through a cascaded evidence flow: the detector gates the subsequent localization and prompting process, the localizer converts tamper responses into grounding boxes, and the reasoner is trained to synthesize the detector decision and localized evidence into the final report. As a key part of our method, we introduce iterative difficulty-aware mining to improve localization quality and apply report-mask consistency post-processing to align report grounding with predicted masks. On the official hidden test set, our system achieves a final score of 0.638 and ranks second in the challenge, validating the effectiveness of the proposed evidence-guided system. The code is available at https://github.com/peifengLiu42/ACMMM26-evidence-guided-detector-localizer-reasoner-system.
Read the full abstract
Information Theory
New binary optimal LCD codes using heuristic embedding
In this paper, we investigate the construction of binary optimal LCD codes through short LCD embeddings. For this purpose, we design heuristic frameworks based on a greedy algorithm. We explore the search spaces of LCD embeddings using the fact that an invertible matrix together with an arbitrary matrix yields an LCD embedding. We therefore use elementary row operations on the invertible block and single entry-flips on the arbitrary block as local moves in a greedy algorithm. Using this method, we have found $14$ optimal new LCD codes with dimensions 7 and 8 for lengths from 55 to 201.
Read the full abstract
Artificial Intelligence
READY or Not: Reliable Enterprise Agent Deployment
An AI agent can perform well on benchmarks and still be unsuitable for deployment. Existing AI-agent benchmarks measure whether an agent can complete realistic professional work, whereas enterprise deployment asks a different question: whether an agent can meet a required reliability level, under acceptable human oversight, and at tolerable cost. We introduce Reliable Enterprise Agent Deployment (READY), a framework for qualifying AI agents for deployment on enterprise workflows. READY preserves each workflow's own definition of successful execution while applying a common qualification procedure. Given an agent, a workflow, and a class of candidate oversight policies, READY measures the reliability and operating cost of the human-AI system, selects the minimum-cost policy that satisfies a specified reliability target, and statistically qualifies it on held-out cases. The resulting deployment profile characterizes the supported operating point: reliability, human-oversight burden, and cost. READY is implemented as an open testbed that decouples workflow specification, execution, evaluation, and qualification, and runs on existing agent-evaluation infrastructure. In an end-to-end clinical-audit case study spanning 16 agent systems and 750 cases, READY reveals differences hidden by autonomous performance: two systems separated by only 0.3 points in autonomous accuracy (72.8% vs. 72.5%) require 39.2% versus 29.6% human review, respectively, to qualify at the same 76% reliability target under the evaluated oversight policy. READY thus shifts enterprise agent evaluation from how well can the agent perform the work? to under what conditions, and at what cost, can it be reliably deployed? By making those conditions explicit and statistically testable, READY provides a basis for comparing agent systems, setting oversight requirements, and making evidence-based deployment decisions.
Read the full abstract
Artificial Intelligence
MASkills: Continual Skills Optimization for Multi-Agent LLM Systems
LLM-based multi-agent systems have shown strong performance on complex tasks, yet continual improvement from interaction experience remains challenging. Existing self-reflection methods build experience memories, but memories are mostly hard to invoke, refine, or scale, while agent skills offer a more actionable unit: structured procedural knowledge that specifies when to act, how to act, and which resources or tools to use. We introduce MASkills, a continual learning framework that optimizes multi-agent LLM systems through agent skills. MASkills presents a new agent-optimization pipeline that integrates skill-conditioned credit assignment, hierarchical credit aggregation, and momentum-smoothed optimization, enabling agent skill libraries to evolve through refinement, induction, consolidation, and pruning. Experiments on HotpotQA, LoCoMo, and GAIA demonstrate the effectiveness of MASkills across multiple agentic tasks. Our code is available at https://github.com/DaRL-GenAI/MASkills
Read the full abstract
Machine Learning
Compositional Spectral Prompts for LLM-based Online Time Series Forecasting
To address the sequential and evolving nature of time series, the Online Time Series Forecasting (OTSF) task has been extensively studied in multiple domains. Existing research focuses on adapting to non-stationary environments by employing memory buffer-based retrieval strategies. However, we observe that such frameworks struggle with long-term adaptation and fail to generalize to unseen patterns. To this end, we introduce CoSPOT, an LLM-based online time series forecasting framework that leverages a pre-trained LLM as the backbone online forecaster, motivated by its strong few-shot capabilities. For efficient online adaptation, CoSPOT keeps the LLM frozen and employs compositional spectral prompts grounded in frequency-domain bases to guide the model with the overall distribution of the input, thereby substantially reducing the number of parameters updated during the online phase. Specifically, CoSPOT decomposes time series into frequency bases and composes the corresponding spectral basis prompts according to their amplitudes, allowing unseen patterns to be represented as new combinations of learned basis prompts. Our extensive experiments on real-world datasets demonstrate the superiority and practicality of CoSPOT across challenging online scenarios, including extended online phases and cross-dataset settings with substantial distribution shifts. Our code is available at https://github.com/seungyoon-Choi/CoSPOT.
Read the full abstract
Artificial Intelligence
Beyond Outcome Gaps: Process-Aware Fairness Diagnosis for LLM-based Multi-Agent Decision Systems
LLM-based multi-agent systems (MAS) are increasingly considered for high-stakes decision-making, yet outcome-based fairness audits can miss where risks arise within the decision trajectory. We present SCOPED-Hiring, a process-aware fairness diagnosis pipeline for LLM-based hiring MAS. SCOPED-Hiring constructs controlled resume variants, runs role-based hiring committees, logs over 311K structured decision trajectories, and converts trajectory fields into quantitative fairness signals organized by six diagnostic lenses: final outcome, counterfactual, process, pathway, dynamic, and design effects. SCOPED-Hiring reveals that balanced final hire rates can mask hidden trajectory unfairness in multi-agent decision trajectories: career gaps trigger suspicion, proxy cues shape qualification judgments, and identity cues lead to unequal investigation. Targeted repair guided by these diagnoses reduces total layered burden by 72.3% while shifting the hire rate by only 1.86 pp, showing that process diagnosis can guide effective repair. Project Page: https://scoped-hiring-project-page.vercel.app/
Read the full abstract
Computation and Language
Selective Knowledge Edit Reversal via Gated Singular Vector Shrinkage
Knowledge editing provides an efficient way to update factual knowledge in large language models. However, malicious edits may introduce safety risks, making it necessary to reverse undesirable editing effects. Existing reversal methods for parameter-modifying edits mainly focus on global removal, which may also erase beneficial edits that should be preserved. In this paper, we study selective reversal of edited knowledge, where the goal is to reverse targeted edited facts while preserving the remaining edited facts. Based on the hypothesis that each edit is sparsely encoded within the dominant subspace of the edited matrix, we propose a spectral-based reversal framework that locates edit-sensitive components within the dominant singular subspace of edited weights. Experiments across multiple settings demonstrate the effectiveness of our method in reversing selected edits while preserving unrelated edited facts. These results suggest that different edits are sparsely encoded within dominant singular components and can be separable when the number of edits is moderate, making selective spectral reversal a promising direction for locating edit-specific components and repairing edited language models.
Read the full abstract
Instrumentation and Methods for Astrophysics
BICEP/Keck XXII: Analysis of the South Pole Atmosphere for CMB Observations
The South Pole is among the driest sites on Earth, and the resulting high atmospheric transparency makes it an excellent site for millimeter-wave astronomical observations. Nevertheless, for bolometric surveys of the cosmic microwave background (CMB), fluctuations in line-of-sight water vapor drive variable atmospheric emission, which has a significant impact on survey sensitivity. We have deployed a water vapor radiometer (WVR) to characterize these fluctuations in space and time and to compare atmospheric and CMB signals with other sites. This WVR has been monitoring conditions from the South Pole over the past nine years, co-located and co-observing with the BICEP series of telescopes. In this work, we analyze the performance of the WVR and the characteristics of the South Pole atmosphere and demonstrate coherent wind-driven transport of water vapor structures across the site. We show that fluctuations in CMB detector timestreams are well correlated with fluctuations in atmospheric water vapor, providing independent justification for baseline filtering procedures used to remove these fluctuations and providing a new pathway to detect and mitigate the temperature-to-polarization leakage that currently limits the precision of ground-based CMB polarization measurements.
Read the full abstract
Computation and Language
IDEEA: training-free Input-Dependent stEEring via Activation cluster matching
Steering aligns large language models (LLMs) by injecting a bias into selected activations at inference time, offering a far cheaper alternative to weight-update methods such as supervised fine-tuning or reinforcement learning. However, most existing training-free steering methods are input-independent: a single direction is fitted once and shared across all inputs. This is fundamentally limiting as different inputs occupy different regions of the activation space and admit different optimal steering directions toward the same target concept, much as the gradient with respect to a fixed loss varies from input to input. We close this gap with IDEEA (Input-Dependent stEEring via Activation cluster matching), a training-free framework for input-dependent steering. IDEEA clusters the positive and negative activation supports per attention head, and solves an optimal-matching problem to construct a set of cluster-conditional directions, all about the target concept. At inference time, it picks from this pool of directions and uses the one that best matches the input's own activation for steering. IDEEA aligns the model toward the target concept while preserving the input's original representation, evidence that activations encoding a concept occupy several distinct sub-regions of the representation space rather than a single one. IDEEA improves the truth $\times$ info rate in TruthfulQA by an average of 9.9% (up to 23.5%) over the best input-independent baseline.
Read the full abstract
Computer Vision
Rendering-in-the-Loop: An Execution-Driven Agent for Interactive Web Development
Multimodal large language models have achieved remarkable progress in front-end web development, generating interactive webpages from multimodal references such as screenshots and interaction videos. However, existing work largely emphasizes visual metrics such as aesthetics and layout similarity, while overlooking the more critical validation of interactive functionality. We present RILA, an execution-driven agent that puts browser rendering in the loop, iteratively editing generated code from runtime interaction feedback. RILA introduces an Action Interaction Verification (AIV) module that replays the reference interaction trajectory on the generated webpage to collect grounded execution-aware observations, and an Execution-aware Rendering Score (ERS) that jointly measures interaction correctness and visual fidelity to guide iterative optimization. We further build an execution-verified data synthesis pipeline that produces diverse, high-quality training data, offering gains complementary to inference-time optimization. On IWR-Bench, RILA consistently improves both interaction and visual fidelity across foundation models. Notably, with our training pipeline, RILA lifts the compact Qwen3.5-9B backbone from 40.40% to 57.52%, surpassing far larger one-shot generators, including the 1T-parameter Kimi-K2.6 (55.61%) and the proprietary GPT-5.5 (55.74%).
Read the full abstract
Statistical Methodology
Efficient Screening Designs for Expensive Black-box Models with Qualitative and Quantitative Factors
Computationally expensive black-box models often involve a large number of input factors with complex interactions and varying importance. Experimental design techniques can be used for quickly identifying the important factors, which can make the optimization of a complex computer model or the training of an expensive machine learning model more efficient. Existing screening designs for black-box models focus mainly on continuous factors, with the maximum one-factor-at-a-time (MOFAT) design being a recent example. In this work, we extend the design to incorporate multiple types of factors, including nominal, ordinal, and discrete-numeric. We first identify the properties leading to optimal screening, where qualitative and quantitative factors should be treated differently. Based on these properties, we propose practical algorithms to efficiently construct MOFAT designs for all types of factors. The usefulness of the design is demonstrated by both numerical experiments and an application to hyperparameter tuning in machine learning models.
Read the full abstract
Populations and Evolution
Mutation–selection balance on an infinite trait space: confinement, drift and equilibrium
We study a trait-structured population model incorporating mutation, selection and density-dependent regulation on a countably infinite trait space. The underlying stochastic process is a continuous-time Markov chain in which individuals reproduce at a trait-independent rate, offspring traits are determined by a mutation kernel on $\mathbb Z$, mortality depends on trait, and births are progressively suppressed as the population approaches a fixed population ceiling. Using results for density-dependent Markov population processes with countably many types, we derive a deterministic approximation in the form of an infinite system of nonlinear differential equations. We establish existence and positive invariance of solutions, and investigate the equilibrium structure of the deterministic system. A fundamental distinction emerges between bounded and confining mortality profiles. When mortality remains bounded, mutation may continually transport mass through the trait space and a stationary trait distribution need not exist. In contrast, when mortality increases without bound as the absolute value of the trait index becomes large, the operator $L=D^{-1}P$, where $P$ and $D$ govern mutation and mortality, is compact. By combining compactness with Kreĭn-Rutman theory for compact positive operators on Banach lattices, we show that $L$ has an algebraically simple principal eigenvalue with a strictly positive eigenvector, and we derive a threshold condition for the existence of a non-zero equilibrium. In this regime the equilibrium is unique, and its trait distribution is determined by the principal eigenvector of $L$. Numerical experiments support the theoretical results and illustrate the contrasting behaviours associated with bounded and confining mortality profiles.
Read the full abstract
Machine Learning
TC-Next: Zero-Shot Multimodal Cyclone Forecasting
We present TropicalCycloneNext (TC-Next), a multimodal deep learning model that forecasts tropical cyclone track and intensity at $6-24$h leads by leveraging a foundation model's forecast of atmospheric kinematic and thermodynamic fields, along with GridSat infrared satellite imagery. Trained only on GraphCast forecasts over the Western Pacific (WP), yet reliant only on generic atmospheric variables, TC-Next on GraphCast lowers track error by $15-44\%$ and intensity error by a factor of $3-6$ relative to the rule-based tracker TempestExtremes; applied without retraining to Pangu-Weather and IFS HRES forecasts, it stays ahead on both. Applied zero-shot to the generic weather fields of WeatherNext Cyclones on the 2025 WP season, it attains lower average intensity error at every lead time, and lower or comparable average track error, than that model's specialized direct tracker in a deterministic comparison. Ablations show that the satellite branch lowers track error at every lead time and intensity error at longer leads. TC-Next thus gives a common, independently verifiable basis for comparing the AI forecasts that inform cyclone-warning and climate-adaptation decisions as best available science.
Read the full abstract
Quantum Physics
A Quantum Algorithm for the Radical of a Lie Algebra: Kernel Projection and Conditioning
Every finite-dimensional real or complex Lie algebra has a largest solvable ideal, its radical. In the compact dynamical Lie algebra (DLA) of a closed quantum system, this radical is the center; projecting onto it isolates directions that commute with the supplied algebra. With sparse Lie-bracket data and a known spectral gap, we construct a quantum circuit that approximately encodes this coefficient-space projector. The general construction combines the derived-algebra map with the Killing form, whereas compactness makes the Killing form alone sufficient. This distinction matters for numerical sensitivity, and our main theorem gives its exact law under an invariant-orthonormal basis adapted to the center and semisimple part. In a compact real algebra with a nonzero semisimple part, the condition number of the general operator is the three-halves power of the Killing-form condition number on that part. The compact projector in turn yields a bounded-error test for whether the center is trivial, provided that any nonzero center has at least a stated minimum dimension. This gives a controlled test for internal conserved directions in compact quantum dynamics.
Read the full abstract
Machine Learning
XMerge: Cross-Axis Selection and Reconstructive Layer Merging for LLM Depth Compression
Removing complete transformer layers preserves a standard serving architecture, but existing depth-compression methods can lose substantial quality, and the loss varies unpredictably across models. We introduce XMerge, a post-training method with two components. Cross-axis selection identifies a block with low relative-magnitude and angular hidden-state change, and local boundary reconstruction re-fits the adjacent surviving block to match the original two-block output. XMerge uses no task labels or end-to-end fine-tuning, and it introduces neither architectural changes nor additional inference-time parameters. Across seven Llama and Qwen backbones (0.5B-8B), five published baselines, and three layer-reduction levels, its advantage over baselines is largest at the most aggressive removal: at k=4 it ranks first on six of seven backbones on CORE (a 22-task aggregate) and, separately, on six of seven on MMLU (five of seven on both at once), while avoiding the large perplexity increases of several competing operators. In a task-level bootstrap, the 95% confidence intervals for the three largest CORE margins exclude zero; the remaining margins are consistent with ties. Across the 14 (model, regime) cells it is also the only evaluated operator that never collapses, ranking top-2 in both zero-shot and in-context regimes; on a first calibration probe (one backbone) it is the best-calibrated operator. Ablations show that local reconstruction provides most of the gain, while cross-axis fusion helps when the two selection axes disagree. The additional construction cost is recovered through per-token decode savings after roughly tens of thousands of requests.
Read the full abstract
Multimedia
Transfer Safety Awareness for Cross-Modal Safety Drift in Multimodal Large Language Models
Visual modality enhances the capabilities of multimodal large language models (MLLMs) but also introduces a safety concern: a benign textual query may convey harmful intent when grounded in a visual image. We term this cross-modal safety drift and our pilot studies show that the safety response rate for such requests is substantially lower than that for requests containing explicitly unsafe text. This paper aims to systematically study this issue. First, we conduct an empirical analysis to identify representative unsafe response patterns. Building on these, we interpret model representations and attentions, revealing that visually risky cues receive limited attention and weakly trigger refusal. Motivated by the observation that safety signals from unsafe text processing can be transferred, we propose safety-awareness representation transfer (SRT), a lightweight direction-refinement method that mitigates cross-modal safety drift with a frozen MLLM backbone. Experiments across multiple benchmarks and models show that SRT effectively improves safety in diverse cross-modal settings while preserving utility. Code is available at https://github.com/cucu220123/safety-awareness.
Read the full abstract
Chemical Physics
Biquaternion Algebra with Bilinear Multiplication: A General, Elegant, and Computationally Advantageous Framework for Relativistic Electronic Structure Calculations on CPUs and GPUs
Quaternion algebra provides a natural representation of time-reversal-symmetric matrix structures in relativistic electronic-structure theory, whereas complementary time-reversal-antisymmetric structures extend this representation to complex quaternions, or biquaternions. Here, we explicitly exploit biquaternion algebra for fundamental objects, including operators, kinetically and magnetically balanced basis functions, and their expectation values, within a unified framework encompassing real, complex, and real quaternion subalgebras as special cases. To realize this framework computationally, we have developed HMATLIB, a biquaternion matrix library implemented within the ReSpect package for pure CPU and hybrid CPU/GPU execution. A key development is a bilinear algorithm for matrix-valued biquaternion multiplication that reduces the number of real matrix–matrix multiplications from 64 to 24 compared with the conventional component-wise approach, while the biquaternion representation effectively doubles the maximum accessible matrix dimension under the same 32-bit indexing constraint. Numerical benchmarks on modern CPU and GPU architectures demonstrate that the biquaternion formulation consistently outperforms its isomorphic complex-algebra counterpart. For the largest matrices reported, hybrid CPU/GPU execution accelerates bilinear matrix multiplication by approximately 5-8 times over pure CPU execution and matrix diagonalization by approximately 12 and 61 times relative to CPU oneAPI MKL and NVHPC OpenBLAS, respectively. These results establish biquaternion algebra as a general and computationally advantageous framework for modern relativistic electronic-structure calculations.
Read the full abstract
Combinatorics
Logarithmic basis number of graphs and regular matroids
The basis number $bn(G)$ of a graph $G$ is the minimum edge-congestion of a basis of its cycle space. We prove that every finite $n$-vertex multigraph satisfies $bn(G)=O(\log n)$, resolving, for simple graphs, a question of Bazargani, Biedl, Bose, Maheshwari and Miraftab, subsequently stated as a conjecture by Miraftab, Morin and Yuditsky. The argument also yields the cycle-rank refinement $bn(G)=O(\log \beta(G))$, where $\beta(G)$ is the dimension of the cycle space, and a reduction of Lehner and Miraftab, based on a theorem of Richter and Shank, then gives $bn(G)=O(\log g)$ for graphs of Euler genus $g$. For regular matroids we prove the ground-set bound $bn(M)=O(\log m)$, where $m=|E(M)|$, and logarithmic bounds in both the rank $r(M)$ and the cycle-space dimension $d$. All these orders are best possible.
Read the full abstract
Robotics
Koopman-Based Robust Model Predictive Control for Nonlinear Systems with Stochastic Intermittent Measurements
Intermittent state measurements pose fundamental challenges to model predictive control of constrained nonlinear systems because prediction uncertainty grows during feedback outages and measurement-triggered resets disrupt nominal state propagation, potentially compromising closed-loop stability and recursive feasibility. This paper develops a Koopman-based stochastic MPC framework with probabilistically truncated soft constraints. Specifically, a Lipschitz-constrained deep Koopman model provides a linear latent predictor, enabling computationally efficient online optimization. The intermittent measurement process is modeled as a two-mode discrete-time Markov chain, yielding a unified Markov jump error model for open-loop propagation and measurement-triggered resets. Under numerically verifiable sufficient conditions, the prediction error is shown to be mean-square ultimately bounded, and an explicit uniform second-moment bound is obtained. A distribution-free probabilistic error radius is then constructed for a prescribed confidence level and used to truncate dropout-dependent constraint tightening. An exact-penalty soft-constraint mechanism accommodates reset-induced jumps and prolonged dropouts. Under the stated terminal compatibility and bounded-disturbance conditions, recursive feasibility and mean-square ultimate boundedness of the closed-loop regulation error are established. Numerical simulations on a visual-servoing tracking task corroborate these theoretical results and demonstrate effective tracking under stochastic measurement unavailability.