arXiv subjects

He Sun

Publications and source records attributed to He Sun.

At least 19 recordsLinked to original sources

arXiv2026

Giant magnetocaloric effect at low fields in triangular-lattice NdMgAl$_{11}$O$_{19}$

Magnetic refrigeration in the sub-Kelvin regime requires refrigerant materials to retain a large magnetic entropy at low temperatures by suppressing magnetic ordering. Quantum spin liquids (QSLs), which evade long-range magnetic ordering while retaining strong quantum fluctuations to the lowest temperatures, therefore provide a promising platform for realizing high-performance magnetic refrigerants. Here, we investigate the magnetic ground state and the magnetocaloric effect of the hexaaluminate, NdMgAl$_{11}$O$_{19}$, in which the Nd$^{3+}$ ions form a network of triangular lattices. Magnetic susceptibility and specific heat measurements indicate a magnetically dynamic state down to 50~mK, consistent with a QSL state. Specific heat measurements further reveal substantial magnetic entropy retained below 50~mK. Quasi-adiabatic demagnetization measurements demonstrate a superior cooling performance of NdMgAl$_{11}$O$_{19}$, which can be cooled to 113~mK from 1.9~K by only a small magnetic field change of 2~T. The outstanding refrigeration performance is attributed to the persistent spin fluctuations associated with the QSL-like ground state, together with a large effective \textit{g} factor and the smallness of the exchange interactions along the easy-axis direction. This study demonstrates that frustration, combined with strong spin-orbit coupling and crystal-electric-field effect in the rare earth magnets provides a promising design principle for next-generation cryogenic magnetic refrigerants.

cond-mat.str-el↗

arXiv2026

Imaging-101: Benchmarking LLM Coding Agents on Scientific Computational Imaging

Computational imaging, which recovers hidden signals from indirect, noisy measurements, underpins quantitative discovery across scientific disciplines, yet building a correct reconstruction pipeline demands deep domain expertise and remains laborious even for domain scientists. We introduce Imaging-101, a benchmark of 57 expert-verified computational imaging tasks spanning six scientific domains, each grounded in a peer-reviewed paper and canonicalized into a standardized four-stage pipeline (preprocessing, forward physics modeling, inverse solver, and visualization) Three evaluation tracks (planning, function-level unit tests, and end-to-end reconstruction) probe distinct agent capabilities across the full pipeline. Evaluating seven frontier LLMs uncovers systematic challenges in applying coding agents to computational imaging that go beyond those exposed by general coding benchmarks, spanning algorithm selection, physical convention handling, and pipeline integration. These findings highlight concrete capability gaps and point toward skill-augmented, domain-specialized agents as a practical path to reliable computational imaging assistance.

cs.AI↗

arXiv2026

From Foundation to Application: Improving VLA Models in Practice

Despite recent progress of VLA foundation models, the disparity between laboratory conditions and real-world applications continues to impede their practical implementation. To bridge this gap, we present LingBot-VLA 2.0, which advances LingBot-VLA through improvements in three functional domains. (1) Generalization across tasks and embodiments. Compared to the previous version, we revamp the data processing pipeline and curate around 60,000 hours of data for pretraining, including 50,000 hours of robot trajectories spanning 20 robot configurations and 10,000 hours of egocentric human videos. (2) Expanded action space in addition to dual-arm hardware platforms. In particular, our system accommodates degrees of freedom for the heads, waists, mobile bases, and dexterous hands, thereby empowering the robots to tackle more complex tasks in practical scenarios. (3) Predictive dynamics modeling for improved temporal reasoning. Specifically, we formulate future prediction as a proxy task, facilitated by a video representation model for semantic priors and a depth estimation model for geometric cues. Evaluations on the GM-100 benchmark, conducted in a generalist setting, validate the beneficial impact of these proposed modifications. Furthermore, benefiting from the expanded pretraining data that covers whole-body degrees of freedom, LingBot-VLA-2.0 demonstrates strong cross-embodiment long-horizon mobile manipulation capability across the two robotic platforms.

cs.RO↗

arXiv2026

BC-DIR: Bandit-Controlled Deadline-Aware Incremental Redundancy for QUIC in V2X Networks

Vehicle-to-Everything (V2X) communications require timely and reliable message delivery under highly dynamic wireless conditions. Existing approaches that integrate forward error correction (FEC) into QUIC rely mainly on proactive redundancy and fall back to retransmission once losses exceed the correction capability of the configured code, leading to inefficient recovery under burst loss and unnecessary overhead when network conditions are favorable. This paper presents a Bandit-Controlled Deadline-Aware Incremental Redundancy (BC-DIR) framework for QUIC-based V2X transport. BC-DIR combines rateless coding with a soft decoding deadline and a redundancy margin, enabling repair to be triggered within the available delivery budget while injecting additional repair symbols beyond the immediate deficit to improve recovery under burst loss. A contextual bandit controller further adapts the redundancy configuration online according to end-to-end feedback. We also develop a deadline-constrained reliability analysis under burst loss, showing the advantage of the proposed repair mechanism over conventional retransmission and the existence of an optimal redundancy margin. Monte Carlo simulations validate the analytical results. BC-DIR is implemented in a QUIC-based transport stack and evaluated in Veins/OMNeT++ under both congested urban V2X scenarios and stable network conditions. Experimental results show that, across different traffic congestion levels, BC-DIR improves completion ratio by 10\%--40\% over benchmark schemes in congested V2X scenarios, while under favorable network conditions it can even reduce overhead by about 1\% compared with native QUIC.

cs.NI↗

arXiv2026

Agent-Assisted Side-Channel Attacks on Non-Prefix KV Cache in RAG

Modern Large Language Model (LLM) serving engines increasingly rely on Retrieval-Augmented Generation (RAG) and non-prefix Key-Value (KV) cache fusion to accelerate long-context, multi-tenant inference. While existing KV cache side-channel attacks require strict linear prefix alignment--rendering them ineffective against real-world RAG queries that contain unique, user-specific private prefixes--we uncover a critical class of structural vulnerabilities inherent to chunk-aware memory scheduling. We demonstrate that the deterministic micro-architectural mechanisms used to align and fuse disjoint memory chunks inadvertently leak a continuous "Step-Wave" timing signature. Exploiting this physical observation, we introduce SpliceLeak, the first end-to-end side-channel attack targeting non-prefix KV cache fusion. SpliceLeak executes a systematic two-phase privacy breach: it first structurally fingerprints the exact length of hidden private prompts, and subsequently manipulates boundary collisions to extract exact semantic content token-by-token. Extensive evaluations on production-grade frameworks (vLLM integrated with LMCache) demonstrate that SpliceLeak achieves up to a 100% extraction success rate in bounded-entropy scenarios. Driven by a deterministic +104 ms hardware latency void, the attack requires as few as 63 requests per token, piercing through realistic continuous batching noise. To resolve the inherent conflict between memory deduplication and security, we propose SpliceDefense, a bipartite mitigation framework consisting of Quantized Chunk Padding (QCP) and Constant-Time Boundary Fusion (CTBF). Our evaluations confirm that SpliceDefense effectively flattens the side-channel signal (Delta TTFT ~ 0) with negligible throughput overhead, preserving the critical benefits of global cache sharing.

cs.CR↗

arXiv2026

Vector Linking via Cross-Model Local Isometric Consistency

We study Vector Linking: given two embedding clouds produced by different black-box encoders over partially overlapping datasets, recover cross-model object correspondences using only vectors. Empirically and theoretically, we show that independently trained contrastive encoders exhibit local geometric consistency: short-range distances are approximately preserved up to a scale factor, while long-range distances are not due to model-specific distortion. Building on this, we propose an iterative, reference-based geometric embedding hashing that recovers vector links from a tiny seed set of paired anchors. It represents each vector by distances to sampled paired anchors, proposes candidate links via hash-space matching, and aggregates evidence across views in a Beta-Bernoulli posterior to bootstrap high-confidence links as new anchors. Experiments across multiple benchmarks and embedding model pairs demonstrate accurate and robust linking under varying overlap, seed budgets, and out-of-domain anchors, with applications to vector database integration and cross-model clustering. Code is available at https://github.com/DBgroup-Edinburgh/VecLinking.

cs.AI↗

arXiv2026

Unbiased Diffusion Variational Inversion via Principled Posterior Matching

Existing score-based methods for inverse problems often resort to approximate minimization of the KL divergence between the inversion distribution and the Bayesian posterior. Such an approximation leads to severe mode collapse and unreliable uncertainty quantification. In this paper, we propose Principled Posterior Matching (PPM), a framework that returns to the fundamentals of variational inference, rather than using tricky approximations. Instead of relying on heuristic approximations, we rigorously formulate the exact optimization of the KL divergence via the integration of Fisher divergence. We derive a tractable, equivalent gradient form of this integral, enabling precise optimization without the biases introduced by prior approximations. Our analysis clearly reveals that the mode collapse in previous methods stems directly from this approximation gap. Supported by our theoretical solution, PPM unifies two complementary paradigms: (1) In variational inference, PPM adopts mass-covering divergences that significantly improve the inversion diversity and uncertainty quantification; (2) In amortized inference, it enables the training of an efficient reconstruction network for rapid, single-step reconstruction. Furthermore, our formulation naturally extends to a broader family of divergence measures by generalizing the integral of the Fisher divergence. We validate PPM across challenging computational imaging tasks, including inpainting, super-resolution fluorescent microscopy, and radio interferometric black-hole imaging. In all experiments, PPM achieves superior reconstruction fidelity, faithful multimodal posterior recovery, and well-calibrated uncertainty estimates, establishing a robust framework for scientific imaging.

cs.CV↗

arXiv2026

Hybrid Full Waveform Inversion Assisted by Rytov Approximation for Musculoskeletal Ultrasound Computed Tomography

Ultrasound computed tomography is emerging as a promising safe and accessible modality for soft-tissue medical imaging, with full waveform inversion playing a key role in unlocking its full potential for high-resolution, quantitative reconstructions. Frequency domain full waveform inversion (FDFWI) for reconstructing spatial maps of acoustic properties in the musculoskeletal system is highly sensitive to the quality of low-frequency signals, making the final imaging outcome vulnerable to issues such as inappropriate initial models and strong scatterings related to bones. To address these challenges, we propose a hybrid full waveform inversion (HFWI) algorithm that incorporates a traveltime inversion algorithm based on the generalized Rytov approximation into the FDFWI framework. This hybrid strategy enhances early-stage inversion quality and substantially reduces sensitivity to the initial model, all while maintaining computational efficiency. Importantly, HFWI achieves results comparable to those obtained using well-constructed initial models, without incurring extra computational cost, thus enabling accurate imaging under realistic, bandwidth-limited conditions. In addition, we introduce a near real-time strategy to update first-arrival traveltimes based on forward-scattered phase variations without requiring extra wavefield simulations. Numerical simulations, as well as \textit{in vitro} and \textit{in vivo} experiments confirm the robustness and efficiency of the proposed approach. HFWI also shows promise to extend to more complex scenarios of musculoskeletal parametric reconstruction.

physics.med-ph↗

arXiv2026

TensorHub: Scalable and Elastic Weight Transfer for LLM RL Training

Modern LLM reinforcement learning (RL) workloads require a highly efficient weight transfer system to scale training across heterogeneous computational resources. However, existing weight transfer approaches either fail to provide flexibility for dynamically scaling clusters or incur fundamental data movement overhead, resulting in poor performance. We introduce Reference-Oriented Storage (ROS), a new storage abstraction for RL weight transfer that exploits the highly replicated model weights in place. ROS presents the illusion that certain versions of the model weights are stored and can be fetched on demand. Underneath, ROS does not physically store any copies of the weights; instead, it tracks the workers that hold these weights on GPUs for inference. Upon request, ROS directly uses them to serve reads. We build TensorHub, a production-quality system that extends the ROS idea with topology-optimized transfer, strong consistency, and fault tolerance. Evaluation shows that TensorHub fully saturates RDMA bandwidth and adapts to three distinct rollout workloads with minimal engineering effort. Specifically, TensorHub reduces total GPU stall time by up to 6.7x for standalone rollouts, accelerates weight update for elastic rollout by 4.8x, and cuts cross-datacenter rollout stall time by 19x. TensorHub has been deployed in production to support cutting-edge RL training.

cs.DC↗

arXiv2026

RefracGS: Novel View Synthesis Through Refractive Water Surfaces with 3D Gaussian Ray Tracing

Novel view synthesis (NVS) through non-planar refractive surfaces presents fundamental challenges due to severe, spatially varying optical distortions. While recent representations like NeRF and 3D Gaussian Splatting (3DGS) excel at NVS, their assumption of straight-line ray propagation fails under these conditions, leading to significant artifacts. To overcome this limitation, we introduce RefracGS, a framework that jointly reconstructs the refractive water surface and the scene beneath the interface. Our key insight is to explicitly decouple the refractive boundary from the target objects: the refractive surface is modeled via a neural height field, capturing wave geometry, while the underlying scene is represented as a 3D Gaussian field. We formulate a refraction-aware Gaussian ray tracing approach that accurately computes non-linear ray trajectories using Snell's law and efficiently renders the underlying Gaussian field while backpropagating the loss gradients to the parameterized refractive surface. Through end-to-end joint optimization of both representations, our method ensures high-fidelity NVS and view-consistent surface recovery. Experiments on both synthetic and real-world scenes with complex waves demonstrate that RefracGS outperforms prior refractive methods in visual quality, while achieving 15x faster training and real-time rendering at 200 FPS. The project page for RefracGS is available at https://yimgshao.github.io/refracgs/.

cs.CV↗

arXiv2026

Time-variable Scattered Light in Herbig Disks Observed with Subaru/SCExAO

Using the Subaru Coronagraphic Extreme Adaptive Optics (SCExAO) instrument, we present near-infrared K-band polarimetric imaging of nine Herbig stars selected from a volume-limited sample within 200 pc. We detect the disks around MWC 480, HD 163296, and HD 143006 for the first time with SCExAO, and compare these observations with previous VLT/SPHERE datasets to identify surface-brightness variability. In MWC 480, we resolve two azimuthal brightness dips near the disk minor axis and find evidence that one of them shifted between 2021 and 2022. In HD 163296, we identify an apparent linear azimuthal motion of a localized peak in polarized intensity along the outer ring over a 15-month baseline. The rapid motion of these features relative to the local Keplerian velocity suggests that the observed variability is driven by changing illumination rather than physical material motion. Due to uncertainties in the underlying scattering background, however, we cannot determine the precise physical origin of the variability. No significant disk variability is detected in HD 143006 over a 10-month baseline. We also report the first detection of a protoplanetary disk using the fast-PDI mode on SCExAO, illustrating both the promise and current limitations of this observing mode. Finally, we report non-detections toward HD 144432, HD 56895, PDS 76, HIP 80425, HD 148352, and HIP 81474. All non-detections with Meeus classifications belong to Group II systems and are likely self-shadowed. For these six systems, we measure the system-integrated polarization fraction and angle of linear polarization, providing quantitative constraints on their unresolved circumstellar environments.

astro-ph.EP↗

arXiv2026

HillInfer: Efficient Long-Context LLM Inference on the Edge with Hierarchical KV Eviction using SmartSSD

Deploying Large Language Models (LLMs) on memory-constrained AI Personal Computers (AIPCs) enables low-latency, privacy-preserving inference, but long-context generation is fundamentally bottlenecked by the linearly growing Key-Value (KV) cache. While dynamic KV eviction mitigates this memory wall, existing offloading strategies either trigger crippling PCIe I/O bottlenecks on standard SSDs or suffer from FPGA resource exhaustion by forcing compute-intensive exact attention on a single, weak Computational Storage Drive (CSD). In this paper, we propose HillInfer, a CSD-assisted KV eviction framework that introduces a paradigm shift: offloading strictly lightweight token importance evaluation to a single CSD (e.g., SmartSSD) on AIPCs. To fully capitalize on this lightweight offloading strategy, HillInfer orchestrates a Hierarchical KV Cache Manager (HKM) that leverages temporal locality and dynamic token hit rates to physically partition cache pools, thereby eliminating cross-device I/O thrashing. Additionally, we design an Adaptive Prefetch-based Pipeline (APP) that adaptively balances the evaluation workload between the host CPU and the SmartSSD, effectively masking the heterogeneous straggler effect. Finally, we introduce a CSD-based Evaluation Configuration (CEC) to enable resource-efficient near-data processing on the FPGA. Extensive experiments on a commodity AIPC demonstrate that HillInfer achieves up to an 8.56$\times$ speedup over state-of-the-art baselines, delivering low-latency, I/O-efficient long-context inference without sacrificing model accuracy.

cs.AR↗

arXiv2026

A Survey on Reconfigurable and Movable Antennas for Wireless Communications and Sensing

Reconfigurable antennas (RAs) and movable antennas (MAs) have been recognized as promising technologies to enhance the performance of wireless communication and sensing systems by introducing additional degrees of freedom (DoFs) in tuning antenna radiation and/or placement. This paradigm shift from conventional non-reconfigurable/movable antennas offers tremendous new opportunities for realizing multi-functional, more adaptive, and efficient next-generation wireless networks. In this paper, we provide a comprehensive survey on the fundamentals, architectures, and applications of these two emerging antenna technologies. First, we provide a chronological overview of the parallel historical development of both RA and MA technologies. Next, we review and classify the state-of-the-art hardware architectures for implementing RAs and MAs, followed by a detailed comparison of their distinct mechanisms, performance metrics, and functionalities. Subsequently, we focus on various applications of RAs and MAs in wireless communication systems, analyzing their respective performance advantages and key design considerations such as mode selection, movement optimization, and channel acquisition. We also explore the significant roles of RAs and MAs in advancing wireless sensing and integrated sensing and communication (ISAC). Furthermore, we present numerical performance comparisons to illustrate the distinct characteristics and complementary advantages of RA and MA systems. Finally, we outline key challenges and identify promising future research directions to inspire further innovations in this burgeoning field.

eess.SP↗

arXiv2026

PETS: A Principled Framework Towards Optimal Trajectory Allocation for Efficient Test-Time Self-Consistency

Test-time scaling can improve model performance by aggregating stochastic reasoning trajectories. However, achieving sample-efficient test-time self-consistency under a limited budget remains an open challenge. We introduce PETS (Principled and Efficient Test-TimeSelf-Consistency), which initiates a principled study of trajectory allocation through an optimization framework. Central to our approach is the self-consistency rate, a new measure defined as agreement with the infinite-budget majority vote. This formulation makes sample-efficient test-time allocation theoretically grounded and amenable to rigorous analysis. We study both offline and online settings. In the offline regime, where all questions are known in advance, we connect trajectory allocation to crowdsourcing, a classic and well-developed area, by modeling reasoning traces as workers. This perspective allows us to leverage rich existing theory, yielding theoretical guarantees and an efficient majority-voting-based allocation algorithm. In the online streaming regime, where questions arrive sequentially and allocations must be made on the fly, we propose a novel method inspired by the offline framework. Our approach adapts budgets to question difficulty while preserving strong theoretical guarantees and computational efficiency. Experiments show that PETS consistently outperforms uniform allocation. On GPQA, PETS achieves perfect self-consistency in both settings while reducing the sampling budget by up to 75% (offline) and 55% (online) relative to uniform allocation. Code is available at https://github.com/ZDCSlab/PETS.

cs.LG↗

arXiv2026

Rubrics as an Attack Surface: Stealthy Preference Drift in LLM Judges

Evaluation and alignment pipelines for large language models increasingly rely on LLM-based judges, whose behavior is guided by natural-language rubrics and validated on benchmarks. We identify a previously under-recognized vulnerability in this workflow, which we term Rubric-Induced Preference Drift (RIPD). Even when rubric edits pass benchmark validation, they can still produce systematic and directional shifts in a judge's preferences on target domains. Because rubrics serve as a high-level decision interface, such drift can emerge from seemingly natural, criterion-preserving edits and remain difficult to detect through aggregate benchmark metrics or limited spot-checking. We further show this vulnerability can be exploited through rubric-based preference attacks, in which benchmark-compliant rubric edits steer judgments away from a fixed human or trusted reference on target domains, systematically inducing RIPD and reducing target-domain accuracy up to 9.5% (helpfulness) and 27.9% (harmlessness). When these judgments are used to generate preference labels for downstream post-training, the induced bias propagates through alignment pipelines and becomes internalized in trained policies. This leads to persistent and systematic drift in model behavior. Overall, our findings highlight evaluation rubrics as a sensitive and manipulable control interface, revealing a system-level alignment risk that extends beyond evaluator reliability alone. The code is available at: https://github.com/ZDCSlab/Rubrics-as-an-Attack-Surface. Warning: Certain sections may contain potentially harmful content that may not be appropriate for all readers.

cs.CR↗

arXiv2026

Channel Gain Map Reconstruction Based on Virtual Scatterer Model

This paper proposes an efficient method for modeling and reconstructing the channel gain map (CGM) based on virtual scatterers. Specifically, we develop a virtual scatterer model to characterize the channel power gain distribution in three-dimensional (3D) space, by capturing the multi-path propagation environment structure and exploiting the angular-domain spatial correlation of scatterer response. In this model, the CGM is represented as a function over a set of tunable parameters for virtual scatterers, including their number, positions, and scatterer response coefficients (SRCs), which can be estimated from a limited number of channel power gain measurements at a given set of locations within the region of interest. This new representation offers a flexible and scalable modeling framework for efficient and accurate CGM reconstruction. Furthermore, we propose a progressive estimation algorithm to acquire the scatterers' parameters. In this algorithm, we gradually increase the number of virtual scatterers to balance the computational complexity and estimation accuracy. In addition, by exploiting the spatial correlation of scatterer response, we propose a Gaussian process regression (GPR)-based inference method to predict the SRCs that cannot be directly estimated. Finally, ray-tracing-based simulation results under realistic physical environments validate the effectiveness of the proposed method, demonstrating that it achieves higher reconstruction accuracy compared to conventional CGM estimation approaches.

cs.IT↗

arXiv2026

A Pragmatic VLA Foundation Model

Offering great potential in robotic manipulation, a capable Vision-Language-Action (VLA) foundation model is expected to faithfully generalize across tasks and platforms while ensuring cost efficiency (e.g., data and GPU hours required for adaptation). To this end, we develop LingBot-VLA with around 20,000 hours of real-world data from 9 popular dual-arm robot configurations. Through a systematic assessment on 4 robotic platforms, each completing 100 tasks with 130 post-training episodes per task, our model achieves clear superiority over competitors, showcasing its strong performance and broad generalizability. We have also built an efficient codebase, which delivers a throughput of 261 samples per second with an 8-GPU training setup, representing a 1.5~2.8$\times$ (depending on the relied VLM base model) speedup over existing VLA-oriented codebases. The above features ensure that our model is well-suited for real-world deployment. To advance the field of robot learning, we provide open access to the code, base model, and benchmark data, with a focus on enabling more challenging tasks and promoting sound evaluation standards.

cs.RO↗

arXiv2026

CooperLLM: Cloud-Edge-End Cooperative Federated Fine-tuning for LLMs via ZOO-based Gradient Correction

Large Language Models (LLMs) perform well on many NLP tasks, but fine-tuning them on resource-constrained mobile devices is challenging due to high memory and computation costs, despite growing demands for privacy-preserving personalization. Federated Learning (FL) enables local-data training, yet existing methods either rely on memory-intensive backpropagation or use zeroth-order optimization (ZOO), which avoids backward passes but suffers from slow convergence and degraded accuracy. We propose CooperLLM, a cloud-assisted edge-end cooperative federated fine-tuning framework that combines ZOO on mobile devices with cloud-guided gradient rectification. Mobile clients perform lightweight ZOO updates on private data, while the cloud fine-tunes on auxiliary public data using backpropagation and injects guided perturbations to rectify local updates, improving convergence and accuracy without violating privacy. To address system bottlenecks, CooperLLM introduces pipeline scheduling and adaptive compression to overlap computation and communication and reduce memory usage. Experiments on multiple Transformer models and datasets show that CooperLLM reduces on-device memory by up to $86.4\%$, accelerates convergence by $8.8 \times$, and improves accuracy by up to 10 percentage points over state-of-the-art ZOO-based baselines.

cs.LG↗