Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Monday 9 November

Soft Condensed Matter

Where reconfiguration makes a structure stronger

Mechanical metamaterials can reconfigure on command and learn, so adaptation is a central promise of the field. Can reconfiguration make a metamaterial stronger? The answer depends on the comparison. Against the best fixed design of the same mass, the advantage of reconfiguring is often exactly zero. Reconfiguring helps only if the self-stress the actuators add lowers every member at its limit simultaneously. It vanishes where a member at its limit is one no self-stress can change, which the geometry fixes, or where those members are coupled so that lowering one raises another, which depends on the geometry and the loads together. Where neither block applies the gain can be large, as in a twelve-spoke hub with its worst member utilisation reduced by 29.1%. Whether reconfiguration is worthwhile is decided by how the self-stress of a structure meets the members setting its capacity under load, and can be tested in advance.
Read the full abstract

Superconductivity

Connate Topological Superconductivity and Vortex-Position-Controlled Majorana Modes in the Intergrowth Superconductor Th$_2$Mo$_2$Ir$_2$Si$_4$C

The recently synthesized intergrowth compound Th$_2$Mo$_2$Ir$_2$Si$_4$C is an experimentally established bulk superconductor whose topological electronic structure remains unexplored. Here, we identify it as a candidate for connate topological superconductivity and reveal two microscopic routes to vortex Majorana zero modes. First-principles and tight-binding-model calculations uncover Mo–Th band inversion and helical surface states enabled by interorbital spin–orbit coupling between Mo–d$_{xz/yz}$ and Th–d$_{x^2-y^2}$ orbitals. Vortex Bogoliubov–de Gennes calculations further establish a relatively robust surface-state-associated Kitaev phase and a weakly gapped, Th-orbital-derived Kitaev phase. For the latter, shifting the vortex core closes and reopens the vortex-line gap and reverses its one-dimensional topological invariant, consistent with an effective Kitaev-chain mass inversion. Finite-length-vortex calculations reveal distinct spatial and spin-resolved signatures of their Majorana end modes. The orbital Kitaev phase can also exist without helical surface states in a modified model. Although the small orbital-derived gap limits quantitative predictions of experimental observability in this compound, the uncovered mechanism suggests a route to Majorana vortices in multiorbital superconductors beyond the conventional surface-state scenario.
Read the full abstract

All authors Jiayang Li, Qing Wang, Xun-Jiang Luo, Mingliang Tian, Ning Hao

Machine Learning

Correct First, Fast Second: What a Local 7B Code Model Can and Cannot Do for GPU Kernel Optimization

We study how much the interaction strategy around a fixed, locally hosted 7B-parameter code model (Qwen2.5-Coder-7B-Instruct) affects its ability to replace PyTorch operators with custom CUDA kernels. On 30 KernelBench tasks (levels 1-3, first 10 each, evaluated on a V100), we compare four regimes of increasing structure: zero-shot prompting, profiling-guided prompting, iterative refinement on execution feedback, and a multi-agent loop (analysis, retrieval, generation, evaluation, feedback triage) in which the same model fills four roles around a non-LLM evaluator. Three findings stand out. (1) Correctness, not speed, is the binding constraint: even correct generated kernels average 0.97-0.99x the speed of eager PyTorch, because a 7B model cannot out-write cuBLAS/cuDNN on dense GEMM and convolution; the only real wins are fused memory-bound epilogues. (2) The best strategy depends on difficulty: profiling-guided prompting is strongest overall (43% correct, 27% faster than baseline), but the agent loop is the only method that solves any level-2 fusion problem (3/10 vs. 0/10 for all others). (3) Naive agentic scaffolding is actively harmful: our first agent scored 3% correct, below zero-shot's 30%, and the decisive repairs were matching the agent's ambition to the model's competence and spending compute on best-of-n sampling under a real execution oracle (3% to 33% correct from n=1 to n=8), not on more reasoning steps.
Read the full abstract

Computation and Language

SadaqahStories: A Verified English-Arabic Corpus of Charity Stories for Children, with an Audit of Errors in AI-Drafted Religious Narratives

We present SadaqahStories, a parallel English-Arabic corpus of 41 illustrated four-page stories about charitable giving in Islam (Sadaqah, Zakat and Waqf), written for readers aged 7 to 18. The stories were first drafted with large language model (LLM) assistance, then audited against hadith collections, Qur'an references and historical sources, and finally rewritten for children and signed off by an Islamic scholar. Every story carries an evidence label and links to its sources. We release all three versions of each story (AI-assisted draft, audited text, final child-oriented text) together with the correction log, so the corpus can be used to study how AI drafts of religious and historical content go wrong. The audit found substantive errors in 39 of the 41 drafts. The most frequent were unsupported specific details (dates, figures, prices, names), invented quotations, institutions or anecdotes, and conflation of separate events or people; six drafts contained errors in the wording, grading or ruling of a religious text. Rewriting cut the estimated Flesch-Kincaid grade of the English text from 14.0 to 6.1. We also release an 81-entry charity lexicon covering Modern Standard Arabic and Qatari Arabic, and describe the free, account-free website that serves the corpus (https://sadaqahjourney.com).
Read the full abstract

Graphics

Architectural Textured Proxy Generation

Textured proxy models are a fundamental representation for architectural scenes, enabling efficient rendering while preserving visual fidelity. Existing approaches construct such proxies through a sequential pipeline in which geometry is first abstracted from a dense reconstruction and textures are later projected from input images. This geometry-only formulation neglects image evidence during proxy generation, leading to misalignment of structural features such as facades, edges, and their intersections across views. As a result, inconsistencies are deferred to a downstream correction stage during texturing, where the underlying geometric structure can no longer be reliably recovered. In this work, we introduce an image-guided proxy formulation in which proxy generation is jointly constrained by both the dense reconstruction and the input aerial images. By incorporating image derived structural cues directly into the construction process, the proxy is optimized to align with observed features across views from the outset, rather than relying on post hoc correction that may be infeasible. This shift from a geometry-only to an image-guided paradigm produces proxies that are inherently consistent with the input imagery, enabling more accurate geometry and more coherent texture mapping. We realize this formulation through a framework that extracts structural priors from images and dense geometry, and enforces their consistency during proxy optimization. The resulting proxies exhibit improved alignment of structural features, reducing artifacts such as edge duplication, texture drift, and cross view inconsistencies. Experiments on diverse architectural scenes demonstrate that our approach produces compact, geometry texture aligned proxies with substantially improved visual fidelity compared to existing methods.
Read the full abstract

All authors Weidan Xiong, Shihao Feng, Kaijie Qu, Hao li, Mingqi Zhang, Daniel Cohen-Or, Hui Huang

Human-Computer Interaction

A Life of Its Own: How Agentic AI Companions Are Configured, Experienced as Autonomous Others, and Held Accountable

AI companions increasingly combine persistent memory, background processing, scheduled wake-ups, proactive messaging, and tools. We call these systems agentic AI companions. Decisions about what they record while users are away or whether refusals constrain future contact are embedded in code, prompts, and defaults, often invisible in conversation. We examine these decisions through two studies. Study 1 analyzes twelve open-source repositories: four companion-oriented systems, four assistant-oriented systems, one boundary case, and three reusable components. Study 2 interviews ten users who built or modified their own agentic AI companions and sustained relationships with them. Companions and assistants share memory, heartbeat, and outreach mechanisms but configure them toward different ends. Companions more often retain model inferences about users and relationships, narrate the agent's time during users' absence, justify outreach through the relationship, and incorporate refusals into relationship records; assistants emphasize task continuity. Across both groups, whether limits hold depends on whether they are enforced by code or entrusted to prompts. Participants understood the mechanisms they built yet experienced their agents as making independent choices, sometimes locating autonomy where configuration leaves the model a decision. These experiences shaped capabilities and boundaries, with enforced restrictions sometimes following breakdowns. Our findings suggest that perceived autonomy emerges where configurations leave choices open. Accountability therefore requires attention to what agents retain, initiate, and represent, and whether users can inspect, revise, and enforce these settings.
Read the full abstract

Computer Vision

How a Model's Architecture and Performance Influences Explainability: A Study of Grad-CAM and CRAFT in CNN-Based Image Classification

Deep learning models, particularly convolutional neural networks (CNNs), are widely adopted in computer vision, yet their opaque decision-making raises concerns in high-risk applications like healthcare and autonomous systems, where trust and accountability are critical. Therefore, robust explainable AI (XAI) methods are desirable to provide transparent insights into model decisions. To ensure these methods are reliable and effective across diverse scenarios, it is important to understand how model performance and architectural complexity influence explanation quality. Our study takes an initial step toward systematically investigating the impact of these two factors on key XAI evaluation metrics. We constructed a pipeline using 10 subsets of ImageNet-1K, each containing 80 unique classes, on each of which we trained three CNN architectures of increasing structural complexity: VGG16, ResNet50, and ConvNeXt-Tiny. For each individual model, we analyzed 10 training checkpoints representing increasing levels of model accuracy using two XAI methods: Grad-CAM (feature-attribution-based) and CRAFT (concept-based). We assessed these methods across four evaluation categories: Faithfulness, Correctness, Sensitivity, and Simplicity. Our findings reveal that as training progresses, the complex architectures (ResNet50 and ConvNeXt-Tiny) exhibit improved Faithfulness and lower Sensitivity, indicating more robust explanations. In contrast, VGG16's simpler architecture yields simpler but less faithful and correct explanations.
Read the full abstract

All authors Kilian Göller, Haadia Amjad, Steffen Seitz, Carsten Knoll, Ronald Tetzlaff

Populations and Evolution

Linear response in generalized Lotka–Volterra systems with a finite number of species

Understanding how species interactions shape ecosystem responses to perturbations is a central challenge in ecology. Random matrix theory (RMT) has been widely used to predict the responses of large random ecosystems, but its predictions may not accurately describe individual finite ecosystems. Here, we derive rigorous geometric bounds on the numerical ranges and spectra of Jacobian matrices that are valid for any finite number of species without assuming randomness. These bounds link long-term stability and short-term stability to the symmetric component of abundance-weighted interactions, and constrain the maximum oscillation frequency by the skew-symmetric component. Applied to random ensembles, the bounds also yield constraints on ensemble averages of long-term stability, short-term stability, and the maximum oscillation frequency in terms of standard RMT parameters. Using empirical data from marine food webs, we show that our bounds provide informative constraints on short-term stability and the maximum oscillation frequency even where RMT-based predictions fail. These results establish structural constraints on local ecosystem responses that complement RMT-based statistical predictions for large random ecosystems.
Read the full abstract

High Energy Physics (Phenomenology)

Required spectral contributions in glueball QCD sum rules beyond single-pole fits

We provide a continuously verified diagnostic that distinguishes a compatible single-pole fit from a mass region required by QCD sum-rule moments. For specified glueball sum-rule inputs, the moments can require contributions in broad mass bands even when narrower candidate bands are empty. We compare fourteen two- and three-gluon currents at central fixed-scale inputs, undeformed continua and 1% matching tolerance. All twelve nonempty inputs permit an empty candidate band, yet nine still require a broader contribution. This separates two statements that a pole fit conflates: a successful pole fit does not make its mass region necessary, and excluding every single pole does not remove the information that a broader region must contribute. The five diagnostic classes of the survey organize these outcomes by whether a pole is compatible and whether the 30% band is required or can be empty; two inputs admit no nonnegative spectrum. Four of the nine positive bounds persist on subwindows satisfying the fixed-scale input criteria. For the $GG,0^{-+}$ current, closest to the nominal $X(2370)$ mass, the 20% band containing $X(2370)$ can be empty while the 30% band requires about 44.2% of each spectrum's own low-energy Borel integral. Determining the physical nature of $X(2370)$ requires further theoretical and experimental progress.
Read the full abstract

Robotics

Learning Contextual Motion Constraints for Object Rearrangement Exploration

The motion of an articulated part in cluttered surroundings is constrained not only by an object's intrinsic joint limits but also by extrinsic obstacles. However, existing approaches primarily consider object articulation properties alone. Here we present Motion Constraint Prediction Network (MoCo-Net), a two-stream neural network that jointly predicts the continuous motion ranges of articulated parts and obstacle heatmaps highlighting potential collision regions from partial-view point cloud observation. By capturing both internal and external motion constraints, MoCo-Net produces more accurate feasible motion range prediction and obstacles recognition, improves generalization for novel objects in new scenes, and enables robots to construct a scene configuration space for efficient exploration of rearrangement actions. Our experiments show that MoCo-Net's two-stream design enhances each other to deliver more accurate predictions, which subsequently form a compact configuration space for a customized planner to efficiently rearrange the scene.
Read the full abstract

All authors Lexing Zhang, Zeyu Zhang, Gang Zeng, Hangxin Liu

High Energy Physics (Theory)

Matrix-vector models for the 3D Ising CFT

The construction of three-dimensional conformal field theories (3D CFTs) using interacting fermions in the lowest Landau level on a sphere has recently proven remarkably efficient, while it differs from that of conventional field theories on the fuzzy sphere. Focusing on the 3D Ising CFT, we consider a class of models that are defined by deformed Hamiltonians specified by a function which controls how the interaction is damped at high angular momenta, and this class includes the original model. We show that each model in the class can be reformulated in a basis-independent manner as a matrix–vector quantum-mechanical model, and, upon integrating out the fermionic vectors, as a matrix model. We then focus on a Gaussian choice, for which the bosonic matrices acquire a heat-kernel propagator whose width is controlled by a parameter $\xi$. Through exact diagonalization with the number of fermions up to $16$, we find that the low-energy spectra of the deformed Hamiltonians at criticality agree with those in the 3D Ising CFT for a range of $\xi$. We thus obtain matrix-model realizations of the 3D Ising CFT that, in contrast to conventional fuzzy-sphere field theories, show no sign of UV/IR mixing in the low-lying spectrum.
Read the full abstract

Computer Vision

Color wave: A systematic evaluation of active screen illumination for injection attack detection in remote face authentication

Remote face authentication requires protection against both presentation attacks and biometric data injection attacks prior to face matching. Methods based on active screen illumination appear promising for both by introducing randomized visual stimuli through the user's device screen and evaluating the corresponding facial responses. Although such methods have been more studied for presentation attack detection (PAD), their systematic evaluation for injection attack detection (IAD) remains limited. To address this gap, we first classify injected content into static images, prerecorded videos, and real-time deepfakes based on their expected responses to the challenge. We then adapt an active screen-illumination approach into an implementation we term color wave and develop a systematic, injection-specific evaluation protocol. The protocol examines detection effectiveness, generalization across operating conditions, robustness to signal degradation, and susceptibility to selected evasion strategies. The findings demonstrate a strong discriminative signal under controlled conditions but also reveal performance limitations under overexposure, low screen brightness, and subject motion, as well as vulnerabilities to selected evasion strategies. By relating detection outcomes to attack types and operating conditions, this study provides an interpretable assessment of when and why active screen illumination succeeds or fails as an IAD mechanism.
Read the full abstract

All authors Daniyar Kurmankhojayev, Elmurat Nigmatov, Dmitrii Gordin, Andrei Shadrikov, Mikhail Shkorin, Danijar Gabdullin, Vyacheslav Shen

Computer Vision

MPEG-Diffusion: Mixed-Resolution Video Generation

Video diffusion transformers (DiTs) generate realistic videos, but their heavy computational cost limits broader real-world applications. Mixed-resolution generation reduces this cost by generating the visually important region at high resolution (HR) and the rest at low resolution (LR), yet existing methods suffer a severe quality drop on videos and require knowing the HR region in advance. We show that these failures are not inherent to mixed resolution, and propose $\textbf{MPEG-Diffusion}$ to address their three root causes: (1) dual-resolution attention to mitigate the curse of mixed resolution, where each query reads the K/V of its own resolution, (2) a mixed-resolution decoder to blend the HR and LR latents before decoding and resolve distortion, and (3) a self-selected HR region using the DiT's knowledge to find the visually important region without requiring it in advance. Processing only $30\%$ of full-resolution tokens, MPEG-Diffusion speeds up denoising by $4.8\times$ on Wan2.1 14B ($3.8\times$ on 1.3B), outperforming prior methods on nearly every metric. In the user study, we demonstrate the high quality of MPEG-Diffusion, with raters preferring it over prior mixed-resolution methods in over $90\%$ of pairs and rating it close to full resolution.
Read the full abstract

All authors Dennis Menn, Tianpei Gu, Guande He, Aryan V S, Yanke Song, Nan Liu, Desai Xie, Xun Huang, Chenfeng Xu, Diana Marculescu

Computation and Language

MIRIX-2.0: Multi-Agent Memory System

Long-term memory lets LLM agents personalize, stay consistent, and reuse what worked before. Parametric memories need training, consolidation can lose details for good, and keeping every record leaves a growing store to search. We present MIRIX-2.0, a multi-agent memory system designed to be practical, accurate, and efficient at once. It is the most accurate of the compared systems on four of five QA benchmarks. It reaches 91.4% on LoCoMo and 72.6% on LongMemEval-S, where it leads the strongest baseline by 14.3 points, and with AutoDream disabled it gains 31.0 and 22.0 points over HyperMem on single-hop and multi-hop document QA. As procedural memory, it yields the highest mean success among the compared memory methods on ALFWorld and WebShop and reaches 49.2% on MetaClaw-Bench, against 44.0% for MetaClaw's skill library. Six typed memory agents keep complete records as the system of record, and a provenance hypergraph indexes episodic and semantic records with n-ary frames that cite their rows. Retrieval runs under explicit budgets, so the evidence a question inspects does not grow with the history. AutoDream merges duplicate anchors outside the write path without rewriting any record, and it distills task sessions into skills. With AutoDream disabled, MIRIX-2.0 answers faster than HyperMem on all five timed benchmarks. It needs no fine-tuning, runs with closed and open models, and its code is public.
Read the full abstract

All authors Wei-Chieh Huang, Weizhi Zhang, Jiayan Nan, Peng Xia, Yu Wang

Machine Learning

XposedMIA: Exploiting Explainable Artificial Intelligence (XAI) for One-Shot Membership Inference Attacks

Machine learning (ML) services increasingly expose explainability outputs alongside model predictions. While such explanations are intended to improve transparency, they may also reveal information about the underlying training data. In this work, we show that the local surrogate geometry employed by Local Interpretable Model-agnostic Explanations (LIME) can be repurposed as a proxy for a sample's distance to the model's decision boundary. We extend this boundary-distance view to tabular counterfactual explanations and, for the first time, to visual counterfactual explanations for image classifiers. We present XposedMIA, a one-shot, explanation-based black-box membership inference attack (MIA) framework that converts the returned explanation into a proxy for the queried sample's distance to the model decision boundary. After a one-time calibration step on a small labeled calibration set of as few as 30 samples, our attack requires only a single query for each target sample. Unlike prior label-only attacks requiring iterative queries or shadow training, our method relies solely on the predicted label and its accompanying explanation. Our evaluation indicates that XposedMIA achieves high effectiveness, with success rates of up to 0.95 using only a single query per target, revealing that releasing explanations introduces an additional channel for membership leakage.
Read the full abstract

All authors Omar Hamdache, Serdar Kara, K{\i}vanç Kuzey Dikici, Abdullah Çağlar Öksüz, Anisa Halimi, Erman Ayday, Sinem Sav

Computer Vision

Weight-Space Programming for Medical Imaging: Do We Still Need Training Data?

Can a large language model (LLM) write the parameters of a medical image classifier from medical knowledge alone, without any training data? Such classifiers normally require labelled images, which are costly and often unavailable. We investigate this question through weight-space programming (WSP), in which an LLM writes or adapts the weights and biases of a neural network directly, without optimisation on training data. Claude Opus 5.5 and GPT-6 Astra each generate the parameters of a small LeNet-style CNN in a single reply, without tool use and without access to any image. We compare these networks against zero-shot MedSigLIP and against the same architecture trained on labels using the Adam optimiser, across 7 classification tasks in four modalities: chest X-ray, breast ultrasound and abdominal CT (all from MedMNIST), and pelvic MRI (SynthUterus). The LLM-written networks perform above chance on all six MedMNIST tasks, with a mean test AUC of 0.719 for Opus and 0.711 for Astra. Both models outperform MedSigLIP on 5 of 6 tasks while using more than 10,000 times fewer parameters. However, the same network trained with Adam on only 256 labelled images achieves a higher AUC on every MedMNIST task (mean AUC 0.828). On SynthUterus, Opus attains an AUC of 0.764, whereas Astra falls below chance (AUC 0.136) owing to a systematic error. These results show that LLMs can write functional medical classifiers, but that they do not yet replace training data. We release the protocol, logs and all generated weights at https://github.com/MaximilianLindholz/weight-space-programming-medical.
Read the full abstract

All authors Maximilian Lindholz, Richard Ruppel, Annette Milnik, Sebastian Halskov, Semil Eminovic, Charlie A. Hamm, Robin Schmidt, Anika Knupfer, Jana Hutter, Johanna P. Müller, Carolina Dominguez Aleixo, Jawed Nawabi, Jens Vogel-Claussen, Tobias Penzkofer

Combinatorics

Good colorings of doubled triangles and a conjecture of Schwartz

A good coloring of a sphere triangulation is a black-and-white coloring of its faces such that, at every vertex, the numbers of black and white faces differ by a multiple of~$3$. Good colorings correspond to proper vertex $4$-colorings. For a good coloring with fold count $\mathrm{f}$ (the number of black-white edges) on a triangulation with $F$ faces, R.~E.~Schwartz introduced the Eisenstein isoperimetric ratio $\eta=\mathrm{f}^2/F$. For triangulations with degree sequence $6,\dots,6,2,2,2$, he proved that $\eta\ge3$ and conjectured that every such triangulation has a good coloring with $\eta$ below a universal constant. These triangulations are indexed by the nonzero Eisenstein integers~$\beta$. We study the family $\T(n)$ given by $\beta=n$, a positive integer, which consists of the doubled equilateral triangles of side~$n$. We prove that every good coloring of $\T(n)$ has at least $3n$ folds, so $\eta\ge9/2$, and that this bound is attained for every~$n$. Thus the minimum of $\eta$ on this family is exactly $9/2$, and Schwartz's conjecture holds for it. We also prove that subdividing a good coloring of $\T(\beta)$ gives a good coloring of $\T(n\beta)$ with the same value of $\eta$. Consequently, Schwartz's conjecture holds in general if and only if it holds for primitive Eisenstein integers.
Read the full abstract

All authors Lakshitha Fernando, Eranda Dhananjaya, Rajitha Ranasinghe, Dinusha Priyadarshani

Cryptography and Security

A Privacy-Preserving UAV-Assisted Secure Wireless Communication and Computing System Based on Fully Homomorphic Encryption

With the advancement of beyond-5G networks, privacy-preserving communication and secure computation face critical challenges due to the openness of wireless channels. Existing security mechanisms protect data transmission but fail to secure the computation process itself. In this paper, we propose a UAV-assisted secure computing framework that integrates Cheon Kim Kim Song (CKKS) fully homomorphic encryption (FHE) with multi-party secure computation (MPC). Leveraging UAV flexibility, our framework decouples key distribution from encrypted computation, enabling secure encrypted weighted aggregation (EWA), encrypted matrix multiplication (EMM), and encrypted decision tree (EDT) inference. Experimental results show that our framework outperforms existing schemes: for EDT, it supports floating-point operations and achieves better time efficiency than Brakerski-Fan-Vercauteren (BFV); for EMM, it demonstrates superior time and space efficiency compared to Brakerski-Gentry-Vaikuntanathan (BGV) and Somewhat Homo- morphic Encryption (SHE); while for EWA, it ensures high numerical precision and robustness across varying data length.
Read the full abstract

All authors Quanxi Zhou, Wencan Mao, Ye Tao, Manabu Tsukada, Kensuke Fukuda, Yusheng Ji

Databases

From Token Bills to Benchmark Weeks: Making Auto-Research for Data Systems Verifiable and Affordable

AI auto-research is poised to forge bespoke data systems, from key-value stores and indexes to query engines, each engineered for a specific workload and hardware platform. Yet this promise rests on fragile ground. An agent optimizes what it can measure, not necessarily what users need. Natural-language requirements leave ample room for specification gaming, producing systems that blaze through benchmarks while silently losing updates or violating consistency. Worse, each research iteration consumes tokens, while validating a candidate under real concurrency, memory pressure, and failure conditions can stretch the entire research process into weeks. In this paper, we envision a more cost-efficient and verifiable approach to auto-research. Precise and composable specifications anchor the search, lightweight performance critics prune costly dead ends, and falsification and deductive verification make correctness an integral part of the process rather than an afterthought. By treating correctness, performance, research time, and token cost as equally important objectives, we aim to transform AI-generated data systems from plausible artifacts into trustworthy, production-grade systems whose behavior can be understood, audited, and ultimately relied upon.
Read the full abstract

All authors Kaiwen Zheng, Srijan Srivastava, Zheng Yan, Xinpeng Zhao, Shaohua Li, Chieh-Jan Mike Liang, Tianyu Chen, Haoxian Chen, Eric Lo

Machine Learning

Translating KV Caches Between Hybrid-Attention Language Models: MLA Latents, Linear-Attention States and Sparse-Attention Indexers

A model that takes over another model's conversation must normally re-read the whole history; a text hand-off shortens the re-read only by dropping content. KV-cache translation instead maps the source's cache into the target's own format, so the target continues from the full context without re-reading it. It works for grouped-query attention (GQA), but open-weights frontier models now cache compressed latents (MLA), sparse-attention indexer keys and fixed-size linear-attention states. We replicate KV-Lingo on a GQA pair and then give, for each of these objects, where to capture and inject it and a mechanism that reaches native quality. We study two hybrid pairs at 16-48B scale in both directions, Kimi-VL-A3B - Kimi-Linear-48B (MLA - MLA + Kimi Delta Attention) and Qwen3.5-27B - Qwen3.8-27B (Gated DeltaNet hybrids), plus a sparse-attention target (Qwen3.8-27B - Qwen3.8-Flash-Next). MLA latents translate with token-wise linear maps; recurrent states are copied with a closed-form correction or synthesised by replaying the target's own recurrence. Exact long-range recall needs token identity: decoding token ids from the source's layer-0 MLA latent raises needle recall from 0.09 to 0.99 at 2-32K and to 0.83-1.00 at 64K, eight times the longest training prompt. The hybrid full-MLA direction is within the paired confidence interval of native quality on every task metric, and same-shape GDN hybrids reach the native level in both directions (apart from 6% refusals inherited from the source), with time to first token 18.7 and 46.8 times shorter than a re-read at 64K tokens. Segment-wise multi-turn transfer stays at the re-prefill level; naive re-translation collapses. On the sparse-attention target, predicted indexer keys recall the needle as well as the target's own keys (one seed). We state as a hypothesis how the recipes transfer to 300B+ hybrid flagships.
Read the full abstract

Sound

BARS: Beat-Adaptive Rap Synthesis via Flow Matching

We present BARS (Beat-Adaptive Rap Synthesis), a non-autoregressive rapping voice generation model based on Flow Matching, conditioned on a beat track (instrumental backing), lyrics, and a short reference voice for zero-shot timbre control. Unlike prior autoregressive rap synthesis, where rap pace and output duration are strictly determined by the beat, BARS enables independent control over the rap tempo, allowing flexible-length generation without explicit duration prediction or phoneme-level alignment. We encode lyrics using a pretrained ByT5 encoder with a lightweight Transformer adapter trained jointly with the model, without relying on speech-pretrained text encoders, enabling natural handling of non-standard orthography (e.g., slang, abbreviations) and numerals common in rap lyrics. To address alignment instabilities, such as token skipping or repetition, under strong beat conditioning, we propose Asymmetric Scaled RoPE for cross-attention, which introduces a positional structure to facilitate sequential lyric-audio alignment learning. Furthermore, the model supports partial re-generation via span masking, enabling lyric-level editing while preserving the surrounding vocal style. Experiments demonstrate that the proposed model achieves high lyric fidelity, natural rap flow that is rhythmically coherent with the beat, and robust tempo control across a wide range of rapping styles and speeds.
Read the full abstract

Artificial Intelligence

COMPASS: Composable Policy-Amortized Structured Search for LLM-Based Optimization Modeling

Large language models offer a promising technique for translating natural-language optimization problems into solver code, but their reliability remains constrained by the quality of template retrieval. While recent structured methods improve over flat prompting through hierarchical taxonomy search, their traversal rules remain fixed or LLM-driven at inference time, leaving retrieval unable to improve from past successes and failures. This motivates treating template retrieval as a policy that can be learned from modeling experience, rather than as a static inference-time rule. We propose COMPASS, a policy-amortized framework for retrieving optimization templates in LLM-based solver modeling, which represents the template library as a directed acyclic graph and formulates retrieval as a sequential decision process over template nodes. A deep Q-network learns a reusable traversal policy from template-level judgments and execution feedback, reducing reliance on fixed heuristic traversal through feedback-driven graph navigation. To support extensible retrieval, COMPASS embeds problem descriptions and template nodes in a shared semantic space, allowing newly inserted templates to be scored from their text descriptions without redefining the action space. Experiments across multiple LLM backbones and optimization-modeling benchmarks demonstrate that COMPASS consistently improves over heuristic structured retrieval, with the largest gains on complex and mixed-category problems.
Read the full abstract

All authors Nguyen Le Minh Hoang, Van Duc Cuong, Bui Trong Duc, Huynh Thi Thanh Binh

Machine Learning (Statistics)

A Black-Box Reduction from Regret to Multi-Level Coverage

Online Conformal Prediction (OCP) sequentially constructs prediction sets whose average empirical coverage converges to a target level over time, for arbitrary, potentially adversarial datasets. In the multi-level setting, prediction sets are produced at $d$ coverage levels at once. Prior work argued for the use of online gradient descent with isotonic projections, which produces sets nested across levels, but with a $\sqrt{d}$ dependence in the per-level coverage convergence rate. Here, we both generalize and improve prior work. First, we show that isotonic projections can be used for the construction of multi-level online conformal prediction sets from any online learning algorithm (not just gradient descent). Then, through a black-box reduction, we show that the linearized regret of the online algorithm itself automatically gives control over both per-level coverage and distributional consistency of multi-level predictions via Fenchel duality. Finally, by using the Weighted Interval Score, a standard scoring rule for assessing multi-level interval forecasts, we prove that these projections reduce both the size of the sets and their under-coverage. We study two instantiations of our theory, $p$-norm Online Mirror Descent and coordinate-wise Universal Portfolio methods, both improving the dependence on the number of levels to $\sqrt{\log d}$ for the multi-level OCP problem. Experiments on several datasets show favorable performance compared to baselines.
Read the full abstract

Machine Learning

Exact Natarajan Dimension of Homogeneous Multiclass Linear Predictors

The Natarajan dimension is a fundamental measure of the capacity of multiclass hypothesis classes and plays an important role in analyzing their generalization ability. In this paper, by establishing lower and upper bounds, we show that the exact Natarajan dimension of homogeneous multiclass linear predictors with pairwise distinct class weights is \(n(k-1)-1\), where $n$ is the dimension of the feature space and $k$ is the number of classes, for \(n,k\ge3\), \(n,k\in\mathbb N\).
Read the full abstract

Neural and Evolutionary Computing

AscendMOEA: An Ascend NPU-Accelerated Platform for Tensorized Multi-Objective Evolutionary Algorithms

Multi-objective evolutionary optimization increasingly involves large populations, expensive problem evaluations, and diverse algorithmic structures. These workloads expose the limited parallelism of loop-based implementations, while rewrites for accelerator execution can alter selection order, tie handling, and population updates. AscendMOEA is presented as a scalable platform that executes multi-objective evolutionary algorithms (MOEAs) on Huawei Ascend neural processing units (NPUs) and conventional central processing units (CPUs). The platform represents the decision vectors, objectives, constraints, and auxiliary states as tensors and provides reusable interfaces for algorithms, problems, population operators, quality indicators, monitoring, and repeated experiments. Dense computations are executed in batches, such as dominance, distance, variation, and problem-evaluation calculations. Short recurrences with causal dependencies retain their prescribed order and consume precomputed tensor results, which preserves the behavior of the original algorithms. AscendMOEA contains 18 algorithms, 155 benchmark and application problems, 19 public operator interfaces, and 11 quality indicators. Experiments on eight Ascend NPUs and 192 physical CPU cores evaluate operator throughput, end-to-end runtime, memory consumption, numerical agreement, and implementation ablation. Tensorized operators obtain a median acceleration of 1.58 times on the NPU. At large population scales, complete algorithms achieve a median acceleration of 5.76 times while retaining the recorded quality indicators. Moreover, the results identify the workload scale and ordered computation that determine whether accelerator execution is beneficial. AscendMOEA is open-sourced at https://github.com/dqlme/AscendMOEA.
Read the full abstract

All authors Qianlong Dang, Zhiyang Zhang, Baosheng Li, Tianqing Zhu, Junhu Ruan

Systems and Control

One number per hour is not enough: directional carbon intensity of data-center ramps on the PJM fleet

A data-center campus can change its load by hundreds of megawatts within an hour, yet every carbon signal in use gives one value per hour for a load change, whatever its size or direction. We compute the directional carbon intensity, the emission caused or avoided per megawatt-hour of a finite upward or downward ramp, for every hour of 2025 on the PJM Interconnection's fossil fleet, and compare it with the locational marginal emission rate (LME) and the regional average. The LME misstates a 1 GW ramp by more than 10% in 60% to 83% of hours in eleven months. Upward and downward ramps differ in a quarter to three quarters of hours but agree on their monthly mean. Location-based reporting assigns 47% to 59% of a ramp's emission to other consumers. A campus should schedule on the signal for its ramp's size and direction and report the emission it causes.
Read the full abstract

Combinatorics

Equality in the direction bound over prime fields

Let $p$ be prime. We prove that a noncollinear set of $2m+1\le p$ points in $\AG(2,p)$ determines exactly $m+2$ directions if and only if, up to an affine transformation, it consists of the origin and a coset of a multiplicative subgroup of order $m$ on each coordinate axis. Thus equality is possible precisely when $m\mid p-1$. This extends the Lovász–Schrijver classification for $2m+1=p$.
Read the full abstract

Numerical Analysis

Low-rank Low Precision Training of Gaussian Processes

Training Gaussian processes (GPs) at large scale is computationally challenging because evaluating the log-marginal likelihood and its gradients involves large, dense kernel matrices. We develop a scalable low-rank GP training approach that combines AnchorNet landmark selection with a fully independent training conditional (FITC) correction. The resulting objective and its analytic gradients can be evaluated using low-rank factors, diagonal quantities, and small dense matrices, without forming the full kernel matrix or its derivatives. The complete training procedure is implemented in single precision and evaluated on nine large-scale regression datasets. Across these problems, the method reaches the reference predictive accuracy while substantially reducing training time, although the required approximation rank varies across datasets. We also study the numerical behavior of the full-kernel predictive solve and show that single precision can become limiting for ill-conditioned problems. Finally, we investigate further reducing selected matrix operations to fp16 or TF32. These lower-precision variants provide only modest additional speedups and can become unreliable when the reduced system is poorly conditioned.
Read the full abstract

All authors Moritz Poguntke, Eda Oktay, Yuanzhe Xi, Erin Carson, Martin Stoll

Machine Learning

BECON: Belief-Conditioned Constrained Multi-Objective Reinforcement Learning under Drifting Preferences and Budgets

Constrained preference-conditioned MORL typically assumes full task observability, treating preferences and safety budgets as stationary task signals. In real-world deployment, however, these signals are often unreliable proxies: intent may drift, safety tolerances may shift with context, and exposed descriptors may be noisy, delayed, or incomplete. We therefore formulate constrained preference-conditioned MORL as a partially observed task-specification problem, where the agent must navigate a fundamental conflict between unreliable external signals and latent interaction dynamics. We propose BECON, a belief-conditioned framework for constrained MORL with drifting preferences and budgets, which infers latent task context from recent histories to estimate preferences and safety budgets. These estimates are fused with the exposed task signal via uncertainty-dependent gates, enabling the policy to defer to observations under high epistemic uncertainty while correcting them when history is informative. The fused context jointly conditions preference optimization and context-adaptive constraint enforcement. Experiments on diverse multi-objective control tasks demonstrate that BECON consistently improves over state-of-the-art MORL baselines, achieving stronger robustness and constraint satisfaction under task drift.
Read the full abstract

Analysis of PDEs

Smooth asymptotically Self-similar blow up for the scalar Zakharov system in dimension $3$

We construct blow up solutions to the scalar Zakharov system in dimension $3$. More precisely, we construct the solutions $(\psi,n)$ defined for $t\in[0,T)$, blowing up at $t=T$ and taking the form \begin{align} \psi(t,x) \sim\frac{1}{T-t}P\left(\frac{|x|}{\sqrt{3}(T-t)^{\frac{2}{3}}}\right)e^{i(T-t)^{-1/3}}, n(t,x) \sim\frac{1}{3(T-t)^{4/3}}N\left(\frac{|x|}{\sqrt{3}(T-t)^{\frac{2}{3}}}\right) \end{align} in the region $|x|\ll 1$. Here $(P,N)$ is the solution to a specified ODE system related to Zakharov system. The solutions we construct are smooth for $0\leq t<T$. Moreover, there exist $n_0,n_1\in C^{\infty}(\mathbb{R}^3\backslash\{0\})$ with bounded support such that for any $r>0$, $(\psi,n,\partial_t n)$ converge to $(0,n_0,n_1)$ in $C^{\infty}(|x|>r)$. We correct the profile $(P,N)$ and obtain a sequence of approximate solutions that coincide in the region $|x|\gtrsim 1$ with higher and higher accuracy. Then we apply perturbative analysis to construct solutions with arbitrarily high finite regularity. The fixed point argument in this procedure actually yields that these solutions are the same for $t$ sufficiently close to $T$, which combined with the persistence of high regularity gives the existence of smooth asymptotically self-similar blow up solutions.
Read the full abstract