arXiv Research Digest

Building photorealistic, animatable full-body digital humans remains a longstanding challenge in computer graphics and vision . We propose a novel animatable avatar representation, termed Wavelet-guided Multi-level Spatial Factorized Blendshapes .

Weight quantization has become a standard tool for efficient LLM deployment, especially for local inference, where models are now routinely served at 2-3 bits per parameter . The state of the art is currently split into two sets of methods: scalar quantization techniques, such as GPTQ or AWQ, which are widely deployed but plateau in accuracy at 3-4 bits per parameters . In this paper, we ask whether this gap is fundamental, or whether a carefully optimized scalar

Story Visualization aims to generate a sequence of images that faithfully depicts a textual narrative that preserve character identity, spatial configuration, and stylistic coherence as the narratives unfold . Maintaining such cross-frame consistency has traditionally relied on explicit memory banks, architectural expansion, or auxiliary language models . We introduce ReCap, a lightweight consistency framework that improves character stability and visual fidelity without modifying the base diffusion backbone .

Discrete diffusion language models (dLLMs) have recently emerged as a promising alternative to autoregressive approaches . But existing heuristic sampling strategies remain inefficient: they choose only a small part of tokens to sample at each step . We propose a general sampling order optimization framework that utilize a neural indicator to decide which tokens should be sampled .

Balance-Guided SINDy (BG-SINDy) reformulates $ell_0-constrained sparse regression as a term-level $2,0-regularized problem . Terms are ranked according to their relative contributions to the governing equation balance rather than their absolute coefficient magnitudes . BG-Sindy alternates between least-squares regression and elimination of negligible terms .

Large Language Models have demonstrated exceptional performance across diverse domains but are increasingly constrained by high inference latency . River-LLM introduces a lightweight KV-Shared Exit River that allows the backbone's missing KV cache to be naturally generated and preserved during the exit process .

Large language models commit unrecoverable reasoning errors mid-generation . LPSR achieves $\mathbf{44.0\%}$ on MATH-500 with an 8B model versus $28.8\%$ for standard AR ($+15.2$ pp) at $5.4\times$ lower token cost .

Uniform Discrete Diffusion Model (UDM) has recently emerged as a promising paradigm for discrete generative modeling . However, its integration with reinforcement learning remains largely unexplored . We observe that naively applying GRPO to UDM leads to training instability and marginal performance gains .

Adaptive Conformal Filtering (ACoFi) combines Hamilton-Jacobi reachability-based safety filters with adaptive conformal inference . Under ACoFi, the filter dynamically adjusts its switching criteria based on the observed errors in its predictions of the safety of actions . This gives a soft safety guarantee rather than a hard safety guarantee .

Camera-only 3D object detection has emerged as a cost-effective and scalable alternative to LiDAR for autonomous driving . SemLT3D is a Semantic-Guided Expert Distillation framework designed to enrich the representation space for underrepresented classes through semantic priors .

Multimodal affective computing aims to predict humans' sentiment, emotion, intention, and opinion using language, acoustic, and visual modalities . Current models often learn spurious correlations that harm generalization under distribution shifts or noisy modalities. We propose a causal modality-invariant representation (CmIR) learning framework for robust multimodal learning .

We propose T-REN (Text-aligned Region Encoder Network), an efficient encoder that maps visual data to a compact set of text-aligned region-level representations . The code and model are available at https://://://github.com/savya08/T-EN .

Models from the AlphaFold (AF) family reliably predict one dominant conformation for most well-ordered proteins but struggle to capture biologically relevant alternate states . ConforNets globally modulate AF3 representations, making them reusable across proteins .

The rapid progress of subject-driven text-to-image synthesis, and DreamBooth, has enabled a consent-free deepfake pipeline . An adversary needs only 4-8 publicly available face images to fine-tune a personalized diffusion model and produce photorealistic harmful content .

Large language models (LLMs) are widely used in retrieval-augmented generation (RAG) When retrieved contexts are noisy, incomplete, or heterogeneous, a single generation process often struggles to reconcile evidence effectively . We propose a multi-agent synthesis approach that structures evidence processing into multiple role-specialized agents . MASS-RAG applies distinct agents for evidence summarization, evidence extraction, reasoning over retrieved documents .

Chain-of-Thought (CoT) reasoning has become a powerful driver of trajectory prediction in VLA-based autonomous driving . Yet its autoregressive nature imposes a latency cost that is prohibitive for real-time deployment . Latent CoT methods attempt to close this gap by compressing reasoning into continuous hidden states . We present OneVL (One-step latent reasoning and planning with Vision-Language explanations)

Vision-language models are rapidly increasing in popularity with an ever expanding list of available models . In many cases, these models have parameters in the tens of billions, but in many cases smaller models are necessary . Unfortunately, there is little research in producing light-weight models .

The key idea is to automatically generate optimization rules that enhance the search for optimal solutions . AutoPPA employs an Explore-Evaluate-Induce ($E^2I$) workflow that contrasts and abstracts rules from diverse generated code pairs rather than manually defined prior knowledge .

Diffusion models have achieved outstanding performance in unsupervised industrial anomaly detection (uIAD) by learning a manifold of normal data . OSD-IRF requires only single step diffusion for uIAD, thanks to the property that IRF holds for any neighboring time step in the denoising process .

We present a systematic evaluation of large language model families -- spanning both proprietary cloud APIs and locally-hosted open-source models . Cloud models achieve 77--89\% overall pass rates; the best local model reaches 77\% (Kimi~K2.5~GGUF~Q3, zero-shot engine), matching mid-tier cloud performance .

SynAgent enables scalable and physically plausible cooperative manipulation by leveraging Solo-to-Cooperative Agent Synergy to transfer skills from single-agent interaction to multi-agent human-object-human scenarios .

TurboQuant is a special case of EDEN obtained by fixing EDEN's scalar scale parameter to $S=1$. EDEN supports both biased and unbiased quantization, each optimized by a different $S$ . EDEN is a 1-bit quantizer that EDEN extended to any $b>0$ bits per coordinate . TurboQuant$ is suboptimal in three ways: (1) its . $(b-1)$-bit step uses the subopt

A KL regularization term is proposed that aligns predicted act distributions with corpus-derived transition patterns . Evaluated on a 60-class German counselling taxonomy using 5-fold cross-validation . This improves macro-F1 by 9--42% relative depending on encoder and substantially improves dialogue-flow alignment .

Epidemic forecasting has become an integral part of real-time infectious disease outbreak response . There is limited understanding on performance of these methods for novel outbreaks with limited historical data . We propose IDOBE, a curated collection of epidemiological time series focused on outbreak forecasting .

Large Language Models (LLMs) show promise in lyric-to-melody generation . But models trained with Supervised Fine-Tuning often produce musically implausible melodies . To address this, we propose a novel alignment framework that instills musical knowledge without human annotation .

State-of-the-art Differentially Private (DP) synthetic data generators such as MST and AIM are widely used, yet tightly auditing their privacy guarantees remains challenging . We introduce a Gaussian Differential Privacy (GDP)-based auditing framework that measures privacy via the full false-positive/false-negative tradeoff .

Large Language Model (LLM) agents create a broader attack surface than traditional applications . AgenTEE places the agent runtime, inference engine, and third-party applications into independently attested confidential virtual machines (cVMs) and mediates their interaction through explicit, verifiable communication channels . The system is built on Arm Confidential Compute Architecture (CCA)

Privacy policies are intended to inform users about how software systems collect and handle data, yet they often remain vague or incomplete . We analyzed 1,000 Android apps across multiple categories, generating 86,836,964 log entries .

We present a scalable, data-driven simulation framework for large-scale heating, ventilation, and air conditioning systems . The proposed approach attains multi-fold speedups over high-fidelity simulation while keeping errors low (MAPE below a few percent)

Biologically-informed neural networks (BINNs) are a variant of PINNs that preserve the known differential operator structure (e.g., reaction-diffusion) while learning constitutive terms via trainable neural subnetworks, enforced through soft residual penalties . Existing BINNs studies are limited to $1\mathrm{D}{+}t$ systems and focus on forward prediction, using the governing partial differential equation as a regulariser rather than an explicit identification target

Adversarial training of binary classifiers can be reformulated as regularized risk minimization involving a nonlocal total variation . We establish a characterization of the subdifferential of this total variation using duality techniques . We provide such duality statements in the space of continuous functions vanishing at infinity on proper metric spaces .

We propose a computational framework for replacing the repeated numerical solution of differential Riccati equations in finite-horizon Linear Quadratic Regulator (LQR) problems by a learned operator surrogate . The resulting model enables fast online evaluation of approximate optimal feedbacks across a wide class of systems . The method offers an effective and scalable alternative for parametric and real-time optimal control applications .

Parkinson's disease (PD) is a chronic neurodegenerative disease . Wearable IMU sensors has become a promising gateway for passive monitoring of PD patients . We propose a self-supervised cross-attention encoder that processes bilateral wrist-worn IMU signals .

Security definitions for declassification exist but mostly do not handle higher-order programs . We use logical relations to build a model (and thus security definition) of where declassification occurs . The key insight required for our model is that we must stop enforcing indistinguishability once a relevant declassification has occurred .

MathNet is a high-quality, large-scale, multimodal, and multilingual dataset of Olympiad-level math problems . MathNet spans 47 countries, 17 languages, and two decades of competitions, comprising 30,676 expert-authored problems with solutions across diverse domains .

Modern sequence models are dominated by Transformers, where self-attention mixes information from the visible context in an input-dependent way . We introduce Sessa, a decoder that places attention inside a feedback path, enabling recurrent many-path aggregation within a layer . Empirically, Sessa achieves the strongest performance on our long-context benchmarks .

Proximal Policy Optimization (PPO) has become the predominant algorithm for on-policy reinforcement learning due to its scalability and empirical robustness across domains . There is a significant disconnect between the foundations of trust region methods and the heuristic clipped objective used in PPO . In this paper, we introduce the Bounded Ratio Reinforcement Learning (BRRL) framework .

Large language models have achieved significant reasoning improvements through reinforcement learning with verifiable rewards (RLVR) Yet as model capabilities grow, constructing high-quality reward signals becomes increasingly difficult, making it essential to understand when RLVR can succeed under weaker forms of supervision . We identify reasoning faithfulness, defined as the extent to which intermediate steps logically support the final answer, as the pre-RL property that predicts which regime a model falls into, while output diversity alone is uninformative .

Platonic Representation Hypothesis suggests neural networks trained on different modalities align and converge toward the same representation of reality . If true, this has significant implications for whether modality choice matters at all .

Modern medicine generates vast multimodal data across siloed systems, yet no existing model integrates the full breadth and temporal depth of the clinical record into a unified patient representation . We introduce Apollo, a model trained and evaluated on over three decades of longitudinal hospital records from a major US hospital system .

In this work, we revisit the problem of active sequential prediction-powered mean estimation . At each round one must decide the query probability of the ground-truth label . If the label is not queried, the prediction from a machine learning model is used .

Fully Unsupervised Score Ensembling (FUSE) is a method for improving verification quality by ensembling verifiers without access to ground truth correctness labels . FUSE typically matches or improves upon semi-supervised alternatives in test-time scaling experiments with diverse sets of generator models, verifiers, and benchmarks .

We propose a distributionally robust approach to risk-sensitive estimation of an unknown signal x from an observed signal y . The unknown signal and observation are modeled as random vectors whose joint probability distribution is unknown, but assumed to belong to a given type-2 Wasserstein ball of distributions . The performance of an estimator is measured according to the conditional value-at-risk (CVaR) of the squared estimation error .

Open-weight language models can be rendered unsafe through several distinct interventions . The resulting models may differ substantially in capabilities, behavioral profile, and internal failure mode . We study behavioral and mechanistic properties of jailbroken models across three unsafe routes: harmful supervised fine-tuning (SFT), harmful reinforcement learning with verifiable rewards (RLVR), and refusal-suppressing abliteration .

Reinforcement Learning (RL) enhances LLM reasoning, yet a paradox emerges as models scale . Strong base models saturate standard benchmarks, yielding correct but homogeneous solutions . Constrained Uniform Top-K Sampling (CUTS) flattens the local optimization landscape by sampling uniformly from constrained high-confidence candidates . CUTS improves Pass@1 accuracy on the challenging AIME25 benchmark by up to 15.1% .

This paper proposes a multi-step probabilistic forecasting framework using a single neural-network based model . Key innovations include a new PI loss function based on an extended log-barrier with an adaptive hyperparameter .

Computational Fluid Dynamics (CFD) is central to race-car aerodynamic development . AI-based surrogate models promise to alleviate this bottleneck, but progress has been constrained by limited complexity of public datasets .

This paper is a step-by-step guide to the complete training cycle of a Physics-Informed Neural Network (PINN) It is a topic that existing tutorials and guides typically delegate to automatic differentiation libraries without exposing the underlying algebra . A companion Jupyter/PyTorch notebook reproduces every manual calculation and the full training pipeline, providing mutual validation between hand-derived and machine-computed gradients .

Multi-Scale Reversible Chaos Game Representation (MS-RCGR) uses rational arithmetic and k-mer decomposition to generate scale-invariant features that preserve complete sequence information while enabling diverse analytical approaches . The reversibility property of our encoding ensures no information loss during transformation .

BAR (Branch-Adapt-Route) trains independent domain experts, each through its own mid-training, supervised finetuning, and reinforcement learning pipeline . Unlike retraining approaches that mix all domains, BAR enables updating individual experts independently with linear cost scaling and no degradation to existing domains . BAR achieves an overall score of 49.1 (averaged across 7 evaluation categories)

Social Norms and Trust were the strongest positive EA predictors across cultures . For Germans, EA was directly influenced by Usefulness, Interest and Anger, showing a functional-affective pattern . For Japanese participants, Trust, Surprise and Fear were the direct associational factors, forming a trust-emotion pattern .

Code-generating Artificial Intelligence has gained popularity within both professional and educational programming settings over the past several years . While research and pedagogy are beginning to cope with this change, computing students are left to bear the unforeseen consequences of AI amidst a dearth of empirical evidence .

We study synthesis for obligation properties expressed in LTLFP, the extension of LTLf to infinite traces . Obligation properties are positive Boolean combinations of safety and guarantee properties . They form the second level of the temporal hierarchy of Manna and Pnueli .

AlphaContext is an evolutionary tree-based psychometric context generator for creativity assessment . Creativity has become a core competence in the era of LLMs and human-AI collaboration, underpinning innovation in real-world problem solving . AlphaContext yields an average improvement of 8% over competitive methods across 6 quality metrics .

Omni-Embed-Audio (OEA) is a retrieval-oriented encoder leveraging multimodal LLMs with native audio understanding . OEA achieves comparable text-to-text retrieval performance to state-of-the-art M2D-CLAP, while demonstrating clear advantages in two critical areas .

DiffTSP is a novel discrete diffusion model that treats TSP as a generative task . It adds noise to the KG through a discrete diffusion process . The reverse process gradually recovers the complete KG conditioned on the incomplete graph . Our approach achieves state-of-the-art performance on three public datasets .

TacticGen formulates tactics as sequences of multi-agent movements and interactions conditioned on the game context . Trained with over 3.3 million events and 100 million tracking frames from top-tier leagues .

Large Audio-Language Models (LALMs) have made significant progress in audio understanding, yet they operate as perception-and-answer systems without explicit reasoning processes . Existing methods for enhancing audio reasoning rely on supervised chain-of-thought (CoT) fine-tuning, which is limited by training data quality . We propose Audio-DeepThinker, a framework built on two core ideas .

Conversational agents, such as ChatGPT and Doubao, have become essential daily assistants for billions of users . To further enhance engagement, these systems are evolving from passive responders to proactive companions .

Large language models are increasingly expected to serve as general-purpose agents that interact with external, stateful tool environments . Training robust agents remains limited by the lack of realistic environments and principled mechanisms for life-long learning . We present a self-evolving training arena for advancing general agent intelligence through scalable environments .

Scaffolded Task Design (STaD) framework generates controlled variations of benchmark tasks based on the concept of scaffolding . The approach enables systematic and scalable probing of model behavior by identifying the specific reasoning skill compositions they lack .

Guard models are widely used to detect harmful content in user prompts and LLM responses . However, state-of-the-art guard models rely on terminal-layer representations and overlook the rich safety-relevant features distributed across internal layers . We present SIREN, a lightweight guard model that harnesses these internal features .

Persistent memory is the bottleneck separating stateless chatbots from long-running agentic systems . We present WorldDB, a memory engine built on three commitments: (i) every node is a world -- a container with its own interior subgraph, ontology scope, composed embedding . (ii) nodes are content-addressed and immutable, so any edit produces a new hash at the node and every ancestor, giving a Merkle-style audit trail for free . (

Closed-loop simulation is a core component of autonomous vehicle development, enabling scalable testing, training, and safety validation before real-world deployment . We present Asset Harvester, an image-to-3D model that converts sparse, in-the-wild object observations from real driving logs into simulation-ready assets .

Peptide-protein interactions (PepPIs) are central to cellular regulation and peptide therapeutics, but characterization remains too slow for large-scale screening . ConGA-PepPI uses asymmetric encoding, bidirectional cross-attention, and progressive transfer from pair prediction to binding-site localization .

Silent automation failures pose a critical safety challenge for partially automated vehicles . How to support a driver in silent failure remains underexplored . We found that providing perceptual cues was most effective in enhancing SA, while communicating system intent was superior for building trust .

In self-supervised learning, self-distilled methods have shown impressive performance . However, state-of-the-art methods usually rely on ensembles of complex mechanisms, with many design choices that are not well understood . In this work, we explore the role of self-diffusion within learning dynamics . We show that even this minimal setup can lead to learned representations .

Privacy represents one of the most critical yet underaddressed barriers to AI adoption in mental healthcare . The proposed system extends our prior work on fine-tuned LLM consortiums for psychiatric diagnosis standardization by fundamentally re-architecting the inference pipeline for fully local execution . The platform is designed to assist clinicians with differential diagnosis and evidence-linked symptom mapping .

PLAG is a pseudo-label-guided anomaly generation method designed to enhance tabular anomaly detection . PLAG uses pseudo-anomalies as guidance signals and decoupling the overall anomaly quantification of a sample into an accumulation of feature-level abnormalities. PLAG achieves state-of-the-art performance against eight representative baselines .

In this paper, we present three neural network architectures designed for real-time classification of weather conditions (sunny, rain, snow, fog) from images . These models aim to capture the stylistic elements present in images . Although developed for weather detection, these architectures are also suitable for other appearance-based classification tasks .

Network change validation remains a critical yet predominantly manual, time-consuming, and error-prone process in modern network operations . We present a novel approach that integrates Generative Agentic AI with a multi-functional Network Digital Twin to automate and streamline network change validation workflows .

Qualitative methods are important to use alongside quantitative methods to improve Human-Robot Interaction (HRI) We introduce continuous focus groups, a longitudinal and co-agential method designed to sustain dialogue with assistive care professionals working with children with autism spectrum disorder .

Non-native speakers (NNSs) often encounter speaking difficulties in multilingual communication . We introduce an AI tool with translation for real-time speaking support . It builds a channel for mutual understanding with native speakers (NSs) to mitigate interactional anxiety .

We present BLF (Bayesian Linguistic Forecaster), an agentic system for binary forecasting that achieves state-of-the-art performance on the ForecastBench benchmark . BLF outperforms all the top public methods, including Cassi, GPT-5, Grok~4.20, and Foresight-32B .

LQM is a hierarchical error taxonomy for diagnosing MT errors through six linguistically grounded levels: sociolinguistics, pragmatics, semantics, morphosyntax, orthography, and graphetics . We construct a bidirectional parallel corpus of 3,850 sentences (550 per variety) spanning seven Arabic dialects (Egyptian, Emirati, Jordanian, Mauritanian, Moroccan, Palestinian, and Yemeni) We evaluate six LLMs in a zero-shot

Static Random Access Memory (SRAM) Physically Unclonable Functions (PUFs) make use of intrinsic manufacturing variations in memory cells to derive device-unique responses . Employing such hardware-rooted fingerprints for authentication, this work demonstrates a threshold-based authentication proof of concept for constrained Industrial Internet of Things devices .

Edge computing enables AI inference closer to data sources, reducing latency and bandwidth costs . Orchestrating AI services across the cloud-edge continuum remains challenging due to dynamic workloads and infrastructure variability . We present AIF-Router, an Active Inference--based routing framework that autonomously learns to balance latency, throughput, and resource utilization across multi-tier AI services without offline training .

Floating-point square-root computation is power- and delay-critical operation in edge-AI, signal-processing, and embedded systems . E2AFS achieves the lowest dynamic power (7.63 mW), the shortest critical-path delay (4.639 ns) compared to existing ESAS and CWAHA architectures .

Large Language Models (LLMs) promise to bridge gap through zero-knowledge hardware programming . Users describe circuits in natural language and an LLM compiles them to a hardware intermediate representation (IR) The most user-friendly IRs yield the worst LLM performance .

The paradigm shift towards local and on-device inference under stringent resource constraints is represented by the tiny machine learning (TinyML) domain . The goal of \gls{tml} is to integrate intelligence into tiny, low-cost devices under strict resource, energy, and latency constraints .

Probabilistic Synchronous Parallel (PSP) is a technique in distributed learning systems to reduce synchronization bottlenecks by sampling a subset of participating nodes per round . PSP has a key limitation: it assumes device behavior is static and different devices are independent . This can lead to unfair distributed synchronization, due to highly available nodes dominating training .

Compute-in-memory (CIM) is a promising architecture that addresses this by reducing data movement through integration of computational logic directly into memory . CIMple is designed to overcome limitations in supporting nonlinear operations and various types of transformer models .

Low Earth Orbit mega-constellations extend the cloud-to-edge continuum into space, enabling satellite edge computing . Federated Learning (FL) in this environment is fundamentally energy-constrained due to dynamic inter-satellite connectivity, heterogeneous onboard computing hardware, and strict power budgets . CroSatFL keeps the ground station off the iterative loop by performing all local training and intermediate aggregations on orbit . This sharply reduces repeated use of bandwidth-limited and energy-expensive

MORPH is the first framework that reformulates ZKP kernels to match AI-ASIC execution . MORPH enables TPUv6e8 to achieve up to 10x higher throughput on NTT and comparable throughput on MSM than GZKP .

EcoShift combines online performance prediction with a dynamic-programming-based allocator to distribute reclaimed power across CPU--GPU applications . EcoShift consistently outperforms state-of-the-art policies, achieving up to 6% average performance improvement .

SLO-Guard is a crash-aware autotuner for vLLM serving that treats crashes as first-class observations . It combines a feasible-first Thermal Budget Annealing (TBA) exploration phase with a warm-started Tree-structured Parzen Estimator (TPE) exploitation phase . We additionally contribute a configuration-repair pass, a GPU-aware KV-cache memory guard, and a four-category crash taxonomy .

Sentinel-bench is an 840-inference empirical framework executing a strict intra-model ablation on Qwen-3.5-9B . By toggling latent reasoning across frozen weights, we isolate the impact of inference-time compute against an adversarial Optimism DAO dataset . Autoregressive baseline (System 1) achieved 100% adversarial robustness, 100% juridical consistency, and state finality in under 13 seconds . Conversely, System 2 reasoning introduced

We introduce a framework, Privatar, to offload avatar reconstruction from headset to untrusted devices within the same local network . Privatar's key insight is that domain-specific knowledge of avatar reconstruction enables provably private offloading at minimal cost . Horizontal Partitioning (HP) offloads only low-energy components . HP provides empirical privacy against expression identification attacks .

RISC-V is emerging as a viable platform for automotive-grade embedded computing . Recent ISO 26262 ASIL-D certifications demonstrate readiness for safety-critical deployment in autonomous driving systems . Functional safety in automotive systems is fundamentally a certification problem rather than processor problem .

Large Language Models (LLMs) have revolutionized various artificial intelligence (AI) applications, from natural language processing to code generation . Traditional systems are often unable to meet the computational demands of these models, particularly in training and inference, present significant challenges . This paper explores the role of cloud platforms and distributed systems in supporting the scalability, efficiency, and optimization of LLMs .

cuNNQS-SCI (Selected Configuration Interaction) method stands out as a state-of-the-art technique, recognized for its high accuracy and scalability . However, its application to larger systems is severely constrained by a hybrid CPU-GPU architecture . We introduce a fully GPU-accelerated SCI framework designed to overcome these bottlenecks .

BlockRaFT framework uses RAFT consensus protocol to elect a leader within a cluster of systems . The elected leader coordinates and distributes workloads across follower nodes . Stateless operations are centralized at the leader, while stateful operations are replicated across the cluster .

Neuro-symbolic AI is gaining traction in domains such as large language models, scientific discovery, and autonomous systems . However, deployment is often constrained by high memory demands, diverse computation patterns, and complex hardware requirements . Existing hardware platforms struggle with large on-chip memory overheads, frequent pipeline stalls, limited I/O bandwidth, and inefficient handling of nonlinear operations .

Hardware verification is one of the most challenging stages of the hardware design process . Verification teams aim to maximize design coverage while ensuring correct behavior and alignment with the specification . Spec2Cov is an agentic framework that automatically generates test stimulus directly from design specifications .

Near-Data Processing (NDP) is an area of interest in near-data processing . Hardware designs for such accelerators are appearing, but there lack clean, portable OS abstractions for programming them . We propose a programming model for NDP devices based on virtual processors and inter-process communication channels .

In modern distributed systems, efficient resource allocation is a vital aspect to maintain scalability, reduce operational costs and ensure fast execution . Predictive models for resource usage are essential tools for optimizing allocation and preventing system bottlenecks .

GPUOS uses NVIDIA NVRTC to compile operators at runtime and inject them into the running kernel . This design enables operator updates without restarting the kernel or recompiling the system . Experiments show that GPUOS achieves up to 15.3x speedup over standard PyTorch on workloads dominated by small operations .

EcoSched is an online scheduler that jointly optimizes GPU count selection and application coscheduling to improve workload level efficiency on multi-processor HPC systems . It achieves up to 14.8% energy savings, 30.1% makespan improvement, and 40.4% EDP reduction over baseline schedulers, with modest performance overhead .

Large language models are increasingly deployed as complex agentic systems that scale with task complexity . However, existing infrastructures' scheduling is unaware of the existence of multiple agents, missing opportunities to optimize resource allocation . We propose Hive, a multi-agent infrastructure that enables algorithm- and task-level scaling .

HIVEMIND is a transparent HTTP proxy that applies five OS-inspired scheduling primitives to eliminate failure modes caused by uncoordinated parallel execution . The proxy requires zero modifications to existing agent code and supports Anthropic, OpenAI, and local model APIs via auto-detected provider profiles .

Semantic Tube Prediction leverages representation geometric to regularize LLM hidden-state trajectories toward locally linear geodesics during fine-tuning . The original STP recipe samples random token sub-spans, which is compatible with the base large language model (LLM) training architecture .

The ability of humans for lifelong learning is an inspiration for deep learning methods . In this work we apply Hebbian learning, a biologically inspired learning process, to sound classification . We propose a kernel plasticity approach that selectively modulates network kernels during incremental learning .

Preference optimization is widely used to align large language models with human preferences . Many margin-based objectives suppress the chosen response along with the rejected one . We propose a plug-and-play RC that adaptively rebalances chosen versus rejected updates to satisfy the DB and mitigate likelihood displacement, without redesigning the base objective .

Diffusion models have emerged as powerful tools for a range of vision tasks, including text-guided image generation and editing . In this work, we explore their potential for object grounding in remote sensing imagery . We propose a hybrid pipeline that integrates diffusion-based localization cues with state-of-the-art segmentation models .

Drifting Models train a one-step generator by evolving samples under a kernel-based drift field, avoiding ODE integration at inference . The drift-field iteration admits a locally repulsive regime in a two-particle surrogate . We derive a contraction threshold for the surrogate and show that a linearly-scheduled friction coefficient gives a finite-horizon bound on the error trajectory .

Processing-in-Memory (PIM) architectures offer a promising solution to the memory bottlenecks in data-intensive machine learning . We explore clustering-based vector quantization approaches, which align well with activation characteristics and PIM's internal bandwidth capabilities . We introduce AQPIM, a novel PIM-aware activation quantization framework based on Product Quantization (PQ)

Large-scale dataset distillation requires storing auxiliary soft labels that can be 30-40x larger on ImageNet-1K and 200x larger than condensed images . We identify two fundamental issues: insufficient image diversity and insufficient supervision diversity . We propose Label Pruning and Quantization for Large-Scale Distillation (LPQLD) to address these challenges .

Large language models are increasingly used as planners for robotic systems, yet how safely they plan remains an open question . To evaluate safe planning systematically, we introduce DESPITE, a benchmark of 12,279 tasks spanning physical and normative dangers with fully deterministic validation . Even near-perfect planning ability does not ensure safety: the best-planning model fails to produce a valid plan on only 0.4% of tasks .

Empirical studies of trained models often report a transient regime in which signal is detectable in a finite gradient descent time window before overfitting dominates . We provide an analytically tractable random-matrix model that reproduces this phenomenon in a linear teacher--student setting .

ProtoCLIP is a refinement strategy for CLIP-style VLMs that improves zero-shot discrimination through targeted data curation and distilled anchor alignment . We construct pathology-focused training subsets with curated negative samples to reduce co-occurrence bias . We also introduce a representation-preserving distillation objective to stabilize adaptation .

Smooth functions on graphs have wide applications in manifold and semi-supervised learning . In this paper, we study a bandit problem where the payoffs of arms are smooth on a graph . This framework is suitable for solving online learning problems that involve graphs .

Large language models using chain-of-thought reasoning often waste substantial compute by producing long, incorrect responses . Abstention can mitigate this by withholding outputs unlikely to be correct . We show that abstaining when the value function falls below this reward strictly outperforms natural baselines .

Daily infrastructure management in preparation for disasters is critical for urban resilience . When bridges remain resilient against disaster-induced external forces, access to hospitals, shops, and residences via metapaths can be sustained, maintaining essential urban functions . The heterogeneous graph construction from open data enables redefining bridge roles for disaster scenarios, supporting maintenance budget decision-making .

Fine-tuning Large Language Models typically relies on large quantities of high-quality annotated data . Reinforcement Learning with Verifiable Rewards (RLVR) results lack applicability in many real-world settings . In this work, we present an empirical study of open-source Small Language Model (SLM) performance after RLVR in low data regimes .

Most data-driven ionospheric forecasting models operate on gridded products, which do not preserve the time-varying sampling structure of satellite-based sensing . We instead model the ionosphere as a dynamic graph over Iono pierce points (IPPs) with connectivity that evolves as satellite positions change . This enables prediction on lines of sight that appear only in the forecast horizon .

Recommender systems have advanced markedly over the past decade by transforming each user/item into a dense embedding vector with deep learning models . Unlike traditional ID hashing, BACO is built on the idea of exploiting collaborative signals in user-item interactions for user and item groupings, such that similar users/items share the same embeddings .

Selection bias arises when the probability that an observation enters a dataset depends on variables related to the quantities of interest, leading to systematic distortions in estimation and uncertainty quantification . Classical corrections, such as inverse-probability weighting or explicit likelihood-based models of the selection process, rely on tractable likelihoods, which limits their applicability in complex stochastic models . Here, we develop a bias-aware simulation-based inference framework that explicitly incorporates selection into neural posterior estimation .

The most common cause of dementia is Alzheimer disease, a progressive neurodegenerative disorder affecting older adults . Early symptoms typically include mild memory impairment and reduced ability to acquire new information . Early diagnosis, pharmacological interventions, and supportive care can slow progression and improve quality of life .

Platypoos is a scale-free planning algorithm that adapts to the unknown scale and smoothness of the reward function . We address the problem of planning in an environment with deterministic dynamics and stochastic rewards with discounted returns .

Variational inference (VI) is a central tool in machine learning, used to approximate an intractable target density by optimising over a tractable family of distributions . As the variational family cannot typically represent the target exactly, guarantees on the quality of the resulting approximation are crucial .

In this paper we investigate forecasting coevolving time series that feature intricate dependencies and nonstationary dynamics by using an LLM Large Language Models approach . We propose a novel modeling approach named ContextAware ARLLM CAARL that provides an interpretable framework to decode the contextual dynamics influencing changes in time series .

DiLaR-PINN is a dissipative latent residual PINN designed to learn unmodeled dissipative effects in a physically consistent manner . Structurally, the residual network operates only on unmeasurable (latent) state components and is parameterized in a skew-dissipative form .

BlockEncoding is a foundational technique in modern quantum algorithms, enabling the implementation of non-unitary operations by embedding them into larger unitary matrices . While theoretically powerful and essential for advanced protocols like Quantum Singular Value Transformation (QSVT) and Quantum Signal Processing (QSP) The generation of compilable implementations of block-encodings poses a formidable challenge .

Zeroth-Order optimization presents a promising memory-efficient paradigm for fine-tuning Large Language Models by relying solely on forward passes . However, its practical adoption is severely constrained by slow wall-clock convergence and high estimation variance . We propose AdaLeZO, an Adaptive Layer-wise ZO optimization framework . We argue that the standard uniform exploration strategy is fundamentally flawed as it fails to account for the heterogeneous sensitivity of layers .

Surrogate models based on neural operators could have a lower computational cost than conventional numerical discretization methods . We successfully apply the deep learning approach to the isotropic Allen-Cahn equation and to anisotropic dendritic growth simulation .