arXiv Research Digest

Billion-parameter Vision-Language-Action (VLA) policies have recently shown impressive performance in robotic manipulation . VLA-AD augments teacher-provided 7-DoF action targets with high-level semantic guidance . At test time, the student policy runs independently with neither the VLA teacher nor the VLM required .

Surrogate Neural Architecture Codesign Package is an open-source AutoML framework for hardware-aware neural architecture codesign and end-to-end FPGA deployment . SNAC-Pack runs a multi-objective global search with Optuna and NSGA-II, loading trials to a shared SQLite store . A hardware surrogate model outputs per-trial resource and latency estimates, avoiding the synthesis cost that would otherwise dominate the search loop .

AgriMind is an ensemble deep learning framework that combines three models to achieve high accuracy in plant disease classification.

ITGPT is a Transformer-based large language model designed for timeseries regression that can leverage large volumes of labeled multimodal data for structured data like text.

RecMem is a memory system for LLM agents that stores interactions in a subconscious layer and uses embedding models for retrieval.

This method uses probabilistic chunk masking to reduce gradient computation in reinforcement learning by selecting a small subset of policy phases per trajectory.

CT-AGD (Curvature-Tuned Accelerated Gradient Descent) is a general boosting procedure that accelerates first-order methods by explicitly capturing the local curvature using finite-difference quotients . It has a comparable storage and computational overhead as adaptive gradient methods such as Adam .

This study shows that replacing one layer's map with another's can cause the system to fail to learn, even though both layers are equivalent.

FORGE is a protocol for evolving agent memory in a staged, population-based way that improves evaluation return.

Asteria separates second-order optimization from GPU training by using training hooks to prepare shadow states in advance, allowing expensive inverse-root computations to proceed asynchronously on the host while GPU computations continue.

Large Language Models (LLMs) have demonstrated strong capabilities across diverse NLP applications, such as translation, text generation, and question answering . SGR builds query-specific subgraphs from external knowledge bases and uses their semantic structure to support multi-step inference .

DebiasRAG addresses LLM bias through self-diagnosed bias contexts from the query through retrieval, then reversely produces debiasing contexts to address potential bias.

This method detects machine-generated texts by using metric-based features rather than complex models, addressing the risk of overfitting.

The rise of AI-generated images poses growing challenges for digital authenticity . ReAlign is a novel framework that distills reasoning texts generated by a GRPO-optimized LLM into a lightweight AIGI detector via contrastive learning .

IVGT is a computer vision technique that uses a neural network to learn continuous and coherent 3D geometry from unposed multi-view images.

The accelerating convergence of smart metering, artificial intelligence, and quantum-inspired combinatorial optimisation is reshaping how energy utilities manage physical infrastructure, customer engagement, and environmental accountability .

Argus is an agentic system that uses a Searcher and Navigator to cooperate and assemble evidence for deep research by verifying missing pieces and producing a source-traced final answer.

Fully Open (FO) models are auditable medical AI systems that expose complete training stacks for rigorous validation.

QSurv is a scalable deep learning framework that enables nonparametric continuous-time modeling without relying on time discretization or restrictive distributional assumptions . We propose a training objective based on Gauss-Legendre numerical quadrature, which approximates the cumulative hazard with high-order accuracy . QSurv achieves competitive predictive performance with advantages in instantaneous hazard function estimation .

A benchmark of seven LLM feedback agents for solving propositional logic problems across 10,836 solution--feedback pairs.

This study examines how distributing deliberation tools across a hierarchy affects performance in cyber defense environments.

A case study for how AI coding systems can be used to generate novel scientific hypotheses . We combine a generic coding agent (Google's AntiGravity) with an LLM-driven tree search algorithm to autonomously generate high-efficiency three-dimensional photovoltaic structures .

MAgSeg uses MLLMs to segment agricultural landscapes from satellite imagery without auxiliary vision decoders.

The paper proposes a second-order optimization framework with Multi-Level Variance Correction for multimodal models, addressing modality competition that can destabilize optimization and limit large-batch scaling.

Global AI weather models can produce skillful medium-range forecasts, but their native 0.25 degrees resolution limits direct use for local hazard applications . We introduce SwAIther-Precip, a lead-time-aware downscaling framework that converts coarse-resolution AIFS forecasts into probabilistic km-scale precipitation fields over Switzerland .

Differential privacy changes the effective sample size governing CVaR learning . For tail mass $τ$, the privacy-relevant sample size is not $n$, but $nτ$ For convex Lipschitz learning, modular upper and lower reductions show that CVAR-specific privacy term necessarily scales as $1/2 .

FedEDAuth is a lightweight client authentication framework that detects and filters malicious participants before model aggregation using outlier analysis and micro-cluster behavior.

Federated Learning (FL) has emerged as a promising paradigm for collaborative model training across distributed edge devices while preserving data privacy . Storing this amount of data centrally is challenging due to issues like limited communication, privacy, and regulations . Choosing the right FL architecture depends on the application's needs .

Endogenous tokens are not money - simplified credit

A federated learning framework that separates structural feature unavailability from conventional missingness to enable indirect cross-client knowledge transfer in heterogeneous feature spaces.

This paper proposes a diffusion-based data poisoning framework for federated learning systems.

This paper discusses practical validity conditions for Byzantine-tolerant federated learning, ensuring robust aggregation by bounding the output within the convex hull of honest data.

Website Fingerprinting has traditionally focused on inferring which website a user visits from encrypted traffic metadata such as packet sizes and timing . An adversary can infer a user's persona using only packet-length and inter-arrival-time sequences . We formalize persona fingerprinting under both closed-set and open-world settings .

CTDG transformers fail to focus on critical nodes due to temporal shift weakening attention contrast, suggesting a simple fix of replacing standard attention with differential attention.

PINNs are powerful surrogates for differential equations but are difficult to train due to spectral bias, stiffness, and poor accuracy. GAN-based adversarial training has shown strong results in improving training.

A context-aware entity-relation extraction system for threat intelligence knowledge graphs that leverages hybrid NLP models and domain ontology to organize diverse cybersecurity sources.

The system secures three distinct layers of cloud infrastructure using machine learning models at each layer to detect known attack patterns and distinguish reliable decisions from uncertain outcomes.

A new framework integrates four production-grade capabilities into one system for distribution utilities.

AI can create biases in human communication, which can be amplified through the network and shift collective opinion.

This method embeds the key-dependent perturbation directly into the learned continuous dynamics of a flow matching model, allowing for dynamic-level watermarking.

Predicting magnetic order in materials remains challenging due to noncollinear and incommensurate magnetic structures, which are difficult to determine through experiments or specialized first-principles methods.

LymphNode is a practical defense against unrestricted oracle access in deep neural networks by using plug-and-play watermarking techniques.

This paper introduces a method to analyze language model organization using a technique inspired by aphasias, measuring how the model's functional patterns correlate with symptoms.

We propose a hypothesis-driven approach to modeling mesoscopic dynamics by learning from a mathematically constrained hypothesis class rather than using fixed equations.

This paper examines AI governance techniques for monitoring and auditing AI-enabled products.

Imitating clinical interventions to support pediatric critical care decisions

Bayesian Amnesic Piecewise-Robust SAC unifies Bayesian Online Change Detection (BOCD) with robust ensemble RL . The BAPR operator is a combination of mode-conditional Bellman operators weighted by a frozen belief distribution .

Ransomware recovery in critical manufacturing infrastructure is not only a backup-restoration problem, but a critical-infrastructure continuity and interdependency problem . After ransomware, a plant may remain unable to schedule work, authenticate operators or release product, reconnect OT assets, or coordinate suppliers .

Entropic Autoencoders are a solution to posterior collapse in generative models by learning non-Gaussian, multimodal latent distributions that produce diverse, data-consistent outputs.

This method extends split conformal prediction to include skew-adaptive features, allowing for more accurate regression prediction.

Contemporary AI increasingly participates in attention allocation, reasoning, synthesis, and decision-making, shaping the very cognitive processes through which humans form beliefs, make decisions, and constitute their sense of self . We identify the risks of unstructured delegation: deskilling, automation bias, transfer of epistemic authority, and oracle-style centralization of knowledge .

AI emotional companions face a safety-rapport paradox: restrictive safeguards can damage supportive alliance, while permissive systems risk user harm . SLIP (Staged Layers of Intervention Protocol) is a four-stage graduated methodology deriving interventions from structured qualitative indicators .

Robot wranglers are professionals who design and integrate robots into human spaces to ensure safe and effective interactions.

Designing efficient power delivery hierarchies for AI accelerators is challenging because the rapidly increasing demand for power density requires careful planning to maximize grid capacity.

Generative video models are used in design animation to create structured content with specific motion and layout requirements.

Training data attribution (TDA) for music generation must answer two questions that copyright analysis requires . Existing methods reduce influence to a single scalar, without revealing which musical aspects are dominant in that influence . We propose ARIA framework that decomposes attribution along musical aspects (five for symbolic music, three for audio)

GenShield is a unified framework that detects and corrects AI-generated images by jointly performing explainable detection and artifact correction.

Machine learning (ML) methods have been developed to aid with de novo molecule design . Data on the environmental impacts of chemical compounds are sparse, resulting in low-fidelity ML oracles and unreliable candidate proposals . We present GEMS-an interactive visual analytics tool that enables domain experts to directly collaborate with a genetic algorithm for molecule design. Users can integrate their expert knowledge to guide the evolutionary process by modifying the scoring function and molecule population without programming knowledge .

AR experiences can be made highly reactive through AI-generated content, requiring careful alignment of perceptual modalities for maximum immersion.

LLMs have shown impressive success in program synthesis, discovering programs that surpass prior solutions . Instead of scoring programs after evaluation, we check whether a candidate satisfies a formally defined property . When the property is violated, we stop evaluation early and provide the LLM with a concrete counterexample showing how the program failed . This feedback drastically reduces both the number of program generations and the evaluation cost .

GenLI is a system that combines interest generation, behavior retrieval, and interest fusion to enhance click-through rate prediction in advertising and recommendation systems.

Large Vision-Language Models (LVLMs) have shown significant progress in video understanding, but they face substantial challenges in tasks requiring precise spatiotemporal localization at the instance level . Existing methods primarily rely on text prompts for human-model interaction, but these prompts struggle to provide precise spatial and temporal references . VideoSeeker seamlessly integrates agentic reasoning with instance-level video understanding tasks .

Ada-Diffuser is a causal diffusion model that learns temporal dynamics and latent interactions simultaneously for planning and control.

This study examines whether large language models for legal reasoning are truly reasoning or artifacts of data contamination.

XSearch is a code search framework that uses concept-to-code alignment to explain search results, making the search more transparent and trustworthy.

Constrained latent state modeling provides a unified framework for learning latent representations under competing constraints, addressing the fragmentation in current machine learning approaches.

ToxiAlert-Bench is a large-scale audio dataset containing 30,000+ clips with 20 fine-grained labels for speech toxicity.

This study measured G2G latency and physiological responses under two latency and bitrate manipulations in a fixed-base driving simulator.

An autonomous system using LLM-guided tree search to generate, evaluate, and optimize executable forecasting software for infectious diseases.

ABS is a technological rule enforcement mechanism that translates the rule into an objective determination of distance between pitches and strike zones.

This paper provides an algebraic exposition of the theory of dyadic morality (TDM) using structural causal modeling (SCM) notation to enable neurosymbolic AI systems to compute morality mathematically.

We derive a new objective for discretizing probability paths and use it to create a training-free entropic inference-time scheduler.

ShopGym is an integrated framework for realistic simulation and scalable benchmarking of e-commerce web agents.

This paper develops a sign-separated finite-time error analysis for constant step-size Q-learning . The analysis identifies a max-induced asymmetry in error dynamics . Negative errors admit an optimal-policy lower comparison, the authors say .

Self-supervised pretraining on molecular graphs has emerged as a promising approach for molecular property prediction . Most existing methods operate at a single structural granularity and treat bond information as auxiliary edge attributes rather than as an independent semantic layer . We propose MolCHG, a multi-level self-supervisory pretraining framework built upon a novel Compositional Hierarchical Graph .

Neuromorphic computing offers an energy-efficient alternative to conventional machine learning through event-driven computation . The growing adoption of edge computing has created an increasing need for workloads capable of operating under strict resource and energy constraints .

Scale is a container scheduling and resource allocation framework for serverless edge computing.

This work presents a 3.4x latency reduction in IMC sorting compared to memristor-based approaches, utilizing 6T SRAM for in-memory sorting.

Temporal random walks are widely used to analyze time-stamped interactions in domains such as microservices, finance, and online platforms . Tempest combines a GPU-native dual-index organization over a shared edge store with a hierarchical cooperative scheduler that dispatches walks at thread, warp, or block granularity based on per-step node convergence .

A work-in-progress vision for a unified language and runtime system that allows applications to scale across the edge and cloud.

Autonomous multi-agent systems hit critical reasoning, coordination and scaling bottlenecks as the size and complexity of their tasks grow . APWA facilitates parallel execution by decomposing workflows into non-interfering subproblems that can be processed using independent resources without cross-communication .

Mat2Boundary is a DSL and compiler for boundary computations that models a broad class of boundary-conditions as affine sparse linear operators . This abstraction unifies halo copying, circular and symmetric mappings, zero padding, block-edge synchronization, and user-defined interpolation .

Dynamic Management of Resources middleware enables MPI process malleability in Slurm via a simple API decoupled from scheduler internals . We evaluate this design on the MareNostrum~5 supercomputer .

We study graph exploration in dynamic graphs using mobile agents and show that exploration is impossible if the number of agents exceeds a specific threshold.

Edge-AI deployment is bottlenecked by data-movement energy . Pair event-driven vision sensors with in-memory analog compute could lift that ceiling by orders of magnitude .

MERSEM uses evolutionary search and reinforcement learning to optimize graph workload allocation and scheduling in edge-cloud systems.

PipeSD accelerates LLM inference by offloading cloud workloads and enabling offline robustness and privacy enhancement through cloud-edge collaborative deployment with speculative decoding.

We study distributed zero-knowledge proofs, introduced by Bick, Kol, and Oshman (SODA 2022) While distributed interactive proofs have advanced rapidly, distributed interactive proof techniques remain limited and mostly problem-specific . Our main contribution is a distributed zero knowledge implementation of Sumcheck . For non-k-colorability, we obtain an $O(n)$-round distributed statistical zero knowledge proof deciding whether a graph is not k-colorable, for any constant k,

This method replaces normalization layers with scalar approximations to bypass global reduction bottlenecks in Vision Transformers for edge deployment.

The constitution assigns, per amendable component including itself, a metric space, aggregation rule, and supermajority threshold . Amendments proceed by members voting with their ideal elements, followed by members submitting public proposals carrying supermajority public support under the revealed votes . Public proposals can be sourced from deliberation among members, vote aggregation, or AI mediation .

EMA adapts system models to handle diverse, long-running, and dynamic network environments, making it suitable for resource management and simulation tasks.

Training LLMs on private data allows for better domain-specific knowledge and real-world utility, addressing a major challenge in secure AI systems.

Rescaled Asynchronous SGD is a distributed optimization method that corrects bias toward a frequency-weighted average by rescaling worker-specific stepsizes, ensuring each worker contributes the same aggregate learning rate over a cycle.

Generative recommendation (GR) has emerged as a promising paradigm that replaces fragmented, scenario-specific architectures with unified Transformer-based models . But deploying GR at scale on Ascend NPUs faces fundamental system-level challenges .

Swarm Network-as-a-Service uses drones to provide on-demand connectivity at scale, treating drone interactions as composable services.

This paper presents an LLM-empowered workflow for RISC-V supply chain analysis . The proposed approach addresses the challenges of heterogeneous and unstructured supply chain data by leveraging LLMs for textual understanding and VLMs for extracting information from visual artifacts such as diagrams and tables .

Nature-inspired computing approaches have gained significant attention for solving difficult optimization problems . SACHI is an all-digital Ising architecture implemented by repurposing the L1 cache of a CPU using SRAM-based processing-in-memory techniques .

Spiking Neural Networks (SNNs) are capable of mimicking the spike-based data processing typical of biological neurons . The spread of digital neuromorphic hardware is slowed down by the prohibitive costs that the silicon tape out of circuits brings . FPGAs could represent a viable alternative, offering a flexible and cost-effective platform .

LCLs are a key model in distributed computing that describe valid solutions to graph problems using constant-radius neighborhoods.

Charm++ is a parallel programming system which has demonstrated the utility of overdecomposition for many applications and in multiple contexts . The emergence of GPGPUs as a dominant compute component has created some real and perceived challenges for this paradigm .

FocalPolicy is a foresight-aware visuomotor policy that combines Frequency-Optimized Chunking with Locally Anchored flow matching . It combines time-domain alignment within the proximal actions while regularizing frequency-domain structure over multiple future action chunks to improve cross-chunk coherence .

This tool detects subtle differences in dataset probability distributions by identifying localised density anomalies in high-dimensional feature spaces.

This system can manipulate process behaviour without creating obvious devices outliers, which is a form of logic-layer deception that can preserve numerically plausible measurements while breaking expected cause-and-effect relationships.

LLQR+SAM combines sharpness-aware minimization with a learned preconditioner to address the issue of ignoring loss geometry.

Multi-Fidelity Flow Matching is a refinement framework that uses cascading refinement to improve parametric PDE solutions by conditioning on data-driven source distributions.

MIND decouples model-induced label noise using Latent Manifold Disentanglement to separate noise components.

Looped SSMs are a recurrent model family that can be improved by repeating blocks across layers and reshaping input data.

This research explains how different model outputs are graded by examining internal mechanisms using position-aware edge attribution patching.

This research uses Q-learning to train an agent to recover plumes in turbulent flows, combining insect behaviors to improve recovery.

Shapley Neuron Valuation quantifies Neuron importance in continual learning by freezing important neurons while keeping others plastic.

A diffusion-based sampler for stipple patterns that satisfies a learned local point-distribution prior and a continuous image-defined capacity constraint.

Continual learning of domain-invariant representations addresses the problem of shortcut learning by training models to learn domain-invariant representations

GOMA is a structure-driven multimodal alignment framework that decouples three key design choices: where messages flow, how multimodal evidence propagates, and smoothing depth.

This system uses autonomous spiking dynamics in clockless digital circuits to create networks of interacting neurons with configurable weights, enabling scalable neuromorphic computing.

Neural network surrogate models have emerged as a promising approach to model solution fields for boundary value problems encountered in physical modeling . Stochastic problems represent an area of interest because of the potential to significantly reduce the repeated evaluation of expensive forward models via traditional numerical solvers when conducting parametric analysis .

We propose a variational quantum classifier operating on high dimensional deep representations via amplitude encoding, stabilized by a learnable classical pre encoding layer . Model reliability is assessed using SAFE-AI metrics derived from the Cramer von Mises divergence .

Machine learning systems increasingly make life-changing decisions about individuals . How can individuals respond to negative decisions made by these opaque systems? We propose an operational definition of contestability as a complement to recourse .

Model-Based Reinforcement Learning (MBRL) leverages latent imagination for sample efficiency, yet remains constrained by Historical Tethering . We propose Mind Dreamer (MD), a framework that operationalizes Active Latent Intervention (ALI) to transcend Markovian continuity . MD reformulates discovery as the minimization of a global Relay Manifold Expected Free Energy (R-EFE)

This paper proposes a framework for training Variational Autoregressive Networks with probability priors, enhancing performance by incorporating physical priors.

Deep Reinforcement Learning (DRL) for quadrotor flight control typically relies on Domain Randomization for sim-to-real transfer . We propose a novel adaptive control architecture that actively perceives and reacts to instantaneous perturbations . Real-world validations on a Crazyflie micro-quadrotor demonstrate that our adaptive controller significantly outperforms baselines .

Testing properties of graphs in graphical models using covariance queries is a specialized statistical problem that helps researchers analyze complex systems by examining relationships between variables.

This paper proposes a simple information-theoretic classification method based on entropy to identify active dimensions in variational models.

We propose a novel definition of model exploitation in reinforcement learning, comparing it to reward hacking to characterize exploitable models.

Connectivity-Constrained and Retrieval-Enhanced (CCRE) deep learning framework proposes a deep-learning framework . CCRE achieves a 72.3% first-destination accuracy and a 61.4% average three-step accuracy, outperforming baselines like CatBoost and LSTM .

Parameter-efficient fine-tuning (PEFT) has emerged as a critical technique for adapting large-scale foundation models across natural language processing and computer vision . We introduce a novel PEFT method that constructs orthogonal transformations through low-rank skew-symmetric matrices and compositional rotation chains . We propose an approximation scheme that enables fully parallel computation of compositional rotations .