arXiv subjects

Pengfei Yang

Publications and source records attributed to Pengfei Yang.

At least 19 recordsLinked to original sources

arXiv2026

LUMIN: Lightweight Universal Manufacturing Inspection Network for Anomaly Detection

Industrial anomaly detection faces two engineering bottlenecks: memory bank construction latency and inference efficiency. Traditional sampling algorithms (Farthest Point Sampling, K-Means, etc.) rely on numerous backbone forward passes and iterative distance computations, with construction times ranging from minutes to hours; heavy computation components such as multi-scale feature extraction struggle to meet the millisecond-level real-time requirements of production lines. This paper focuses on sampling efficiency and inference optimization for industrial deployment with two core contributions: (1) PSP (Plugin Sampler Pipeline)---a four-stage adaptive memory bank sampling pipeline based on 18-dimensional pixel metadata and five complementary visual plugins. PSP completes all sampling with zero backbone forward passes; coarse filtering is sub-second numerical sorting, and metadata extraction is a one-time offline cost. PSP supports progressive deployment and incremental updates. (2) Two engineering optimization strategies---parallel memory bank similarity computation (reducing inference memory and latency by over 95\%) and stratified pixel sampling for large-scale evaluation (reducing computation time by 20$\times$ while keeping metrics stable). As a vehicle for validation, we introduce LUMIN (Lightweight Universal Manufacturing Inspection Network) with extreme segmentation-head compression, systematically exploring the accuracy-efficiency frontier against strong baselines. Experiments on five benchmarks demonstrate that PSP matches state-of-the-art sampling accuracy at near-random construction cost (341$\times$ faster than FPS), while inference optimizations reduce evaluation time by 20$\times$ with negligible accuracy loss.

cs.CV↗

arXiv2026

LLM4LLM: Bridging Kernel Benchmarks and Real Deployment via Closed-Loop Agentic Optimization

Large language models have become increasingly capable agents for low-level code and kernel optimization, but isolated kernel benchmarks provide only a proxy for the deployment behavior that matters in language-model inference. We identify a benchmark-to-deployment gap: candidate kernels that appear correct and fast in standalone harnesses can exhibit different performance, safety, or phase behavior after integration into a real inference workload. We introduce LLM4LLM, a deployment-aware closed-loop optimization framework that starts from a target inference script, extracts phase-aware optimization tasks, searches with an experience-guided episodic agent, and accepts patches through in-model validation. Across ten language-model inference workloads on A100 and H100 GPUs, LLM4LLM improves end-to-end latency for every evaluated model, achieving 3.91$\times$/6.98$\times$ geometric-mean speedups on A100/H100; as supporting kernel-level evidence, it also attains up to 2.745$\times$ GeoMean speedup on KernelBench Level 2.

cs.AI↗

arXiv2026

ToolHazard: Scaling Adversarial Environments for Security Evaluation and Alignment of LLM-based Agents

Large language model (LLM) agents integrated with external tools are vulnerable to indirect prompt injections embedded in environmental states. However, existing studies largely rely on manually implemented or reused environments, stochastic LLM-based tool simulation, and predefined injection locations, limiting scalable security research across broader domains. To bridge this gap, we propose **ToolHazard**, a scalable adversarial environment synthesis framework that reduces human engineering and supports expansion with additional seed domains and compute. Through an Environment Simulator, an Attacker Agent, and a User Simulator, ToolHazard synthesizes executable stateful environments, discovers viable injection points and generates environment-specific payloads, and constructs state-grounded long-horizon tasks. Based on ToolHazard, we build **ToolHazard-Bench** for stress-testing agents under complex workflows and diverse environmental attacks. Experiments reveal substantial agent vulnerabilities and show that injection timing and placement affect attack effectiveness. Moreover, ToolHazard-generated alignment data improves security on both ToolHazard-Bench and AgentDojo while preserving benign task utility.

cs.CR↗

arXiv2026

Synthesizing Probabilistic Saturating Counters with Differentially Private Formal Guarantees

Branch predictors improve instruction-level parallelism in modern processors and are commonly modeled using saturating counters. However, classical saturating counters are deterministic and thus vulnerable to side-channel attacks: an attacker can manipulate the counter state and infer the branch direction of a victim process. Probabilistic saturating counters (PSCs) have been proposed to mitigate this leakage by randomizing counter updates, but existing evaluations are mainly empirical. In this paper, we give a formal analysis based on differential privacy (DP): we model PSCs and the corresponding Prime+Probe attack strategies as probabilistic Moore machines, derive optimal attack strategies, and quantify the attacker's distinguishing power through DP. Our DP guarantee applies to the PSC primitive under the Prime+Probe observation model; end-to-end security for a full branch predictor under repeated or adaptive attacks is an important direction for future work. We then synthesize parameters for an enhanced PSC that satisfies a target pure DP guarantee. To evaluate utility, we derive the stationary misprediction rate and validate the theoretical predictions on benchmark programs. Compared to deterministic and existing probabilistic saturating counters, the synthesized PSCs provide formal security guarantees while preserving competitive prediction performance.

cs.CR↗

arXiv2025

Liquid Handling of the JUNO Experiment

The Filling, Overflow, and Circulation (FOC) system is a critical subsystem of the Jiangmen Underground Neutrino Observatory (JUNO), responsible for the safe handling of the Liquid Scintillator (LS) and water throughout the detector's commissioning and operational lifetime. This paper details the design and operation of the FOC system, which accomplished the filling of the world's largest LS detector--taking 45 days for water (6.4*10^4 m^3) and 200 days for LS (2.3*10^4 m^3). Throughout water filling, the liquid level difference between the Central Detector and Water Pool was rigorously maintained within safety limits. During LS filling, level control achieved +/-2 cm precision with flow regulation within +/-0.5% of setpoints. An automated control system based on Programmable Logic Controllers and the Experimental Physics and Industrial Control System framework ensured reliable operation. The system preserved LS radiopurity, maintaining 222Rn below 1 mBq/m^3 during filling and achieving 238U/232Th concentrations below 10^-16 g/g. The successful commissioning and operation of the FOC system have established it as an indispensable foundation for the stable long-term operation of the JUNO detector.

physics.ins-det↗

arXiv2025

Multidimensional multiplicative Poisson vertex algebras

In this paper we introduce the notion of multidimensional multiplicative Poisson vertex algebra, the generalization of the notion of multiplicative Poisson vertex algebra to a difference algebra endowed with D commuting shifts. After showing the equivalence of this notion to the notion of Hamiltonian difference operator on a D-dimensional lattice, we characterize scalar local Hamiltonian difference operators up to the order (-2,2) and investigate the bi-Hamiltonian pairs they form.

nlin.SI↗

arXiv2025

Lethe: Layer- and Time-Adaptive KV Cache Pruning for Reasoning-Intensive LLM Serving

Generative reasoning with large language models (LLMs) often involves long decoding sequences, leading to substantial memory and latency overheads from accumulating key-value (KV) caches. While existing KV compression methods primarily focus on reducing prefill memory from long input sequences, they fall short in addressing the dynamic and layer-sensitive nature of long-form generation, which is central to reasoning tasks. We propose Lethe, a dynamic KV cache management framework that introduces adaptivity along both the spatial and temporal dimensions of decoding. Along the spatial dimension, Lethe performs layerwise sparsity-aware allocation, assigning token pruning budgets to each transformer layer based on estimated attention redundancy. Along the temporal dimension, Lethe conducts multi-round token pruning during generation, driven by a Recency-Aware Selective Retention} (RASR) mechanism. RASR extends traditional recency-based heuristics by also considering token relevance derived from evolving attention patterns, enabling informed decisions about which tokens to retain or evict. Empirical results demonstrate that Lethe achieves a favorable balance between efficiency and generation quality across diverse models and tasks, increases throughput by up to 2.56x.

cs.LG↗

arXiv2025

Enhancing Agentic Textual Graph Retrieval with Synthetic Stepwise Supervision

Integrating textual graphs into Large Language Models (LLMs) is promising for complex graph-based QA. However, a key bottleneck is retrieving informative yet compact subgraphs that fit the LLM context. Existing retrievers often struggle, relying either on shallow embedding similarity or costly interactive policies that require excessive supervision. To address these challenges, we introduce an agentic textual graph reasoning framework featuring an LLM-based retriever trained with synthetic stepwise supervision. Rather than relying on final answer rewards which often yield sparse and unstable signals, we optimize the retriever by evaluating each step against offline-extracted golden subgraphs. Our approach distills golden subgraphs via a specialized data synthesis pipeline to formulate dense rewards, facilitating a two-stage training scheme that effectively learns the interactive graph exploration policy. Based on extensive experiments on three common datasets in comparison with seven strong baselines, our approach achieves an average improvement of 15.6% in accuracy and 17.2% in F1 score. The advantage is even higher in more complicated multi-hop reasoning tasks.

cs.CL↗

arXiv2025

GRAIL:Learning to Interact with Large Knowledge Graphs for Retrieval Augmented Reasoning

Large Language Models (LLMs) integrated with Retrieval-Augmented Generation (RAG) techniques have exhibited remarkable performance across a wide range of domains. However, existing RAG approaches primarily operate on unstructured data and demonstrate limited capability in handling structured knowledge such as knowledge graphs. Meanwhile, current graph retrieval methods fundamentally struggle to capture holistic graph structures while simultaneously facing precision control challenges that manifest as either critical information gaps or excessive redundant connections, collectively undermining reasoning performance. To address this challenge, we propose GRAIL: Graph-Retrieval Augmented Interactive Learning, a framework designed to interact with large-scale graphs for retrieval-augmented reasoning. Specifically, GRAIL integrates LLM-guided random exploration with path filtering to establish a data synthesis pipeline, where a fine-grained reasoning trajectory is automatically generated for each task. Based on the synthesized data, we then employ a two-stage training process to learn a policy that dynamically decides the optimal actions at each reasoning step. The overall objective of precision-conciseness balance in graph retrieval is decoupled into fine-grained process-supervised rewards to enhance data efficiency and training stability. In practical deployment, GRAIL adopts an interactive retrieval paradigm, enabling the model to autonomously explore graph paths while dynamically balancing retrieval breadth and precision. Extensive experiments have shown that GRAIL achieves an average accuracy improvement of 21.01% and F1 improvement of 22.43% on three knowledge graph question-answering datasets. Our source code and datasets is available at https://github.com/Changgeww/GRAIL.

cs.AI↗

arXiv2025

Uncertainty-Driven Expert Control: Enhancing the Reliability of Medical Vision-Language Models

The rapid advancements in Vision Language Models (VLMs) have prompted the development of multi-modal medical assistant systems. Despite this progress, current models still have inherent probabilistic uncertainties, often producing erroneous or unverified responses-an issue with serious implications in medical applications. Existing methods aim to enhance the performance of Medical Vision Language Model (MedVLM) by adjusting model structure, fine-tuning with high-quality data, or through preference fine-tuning. However, these training-dependent strategies are costly and still lack sufficient alignment with clinical expertise. To address these issues, we propose an expert-in-the-loop framework named Expert-Controlled Classifier-Free Guidance (Expert-CFG) to align MedVLM with clinical expertise without additional training. This framework introduces an uncertainty estimation strategy to identify unreliable outputs. It then retrieves relevant references to assist experts in highlighting key terms and applies classifier-free guidance to refine the token embeddings of MedVLM, ensuring that the adjusted outputs are correct and align with expert highlights. Evaluations across three medical visual question answering benchmarks demonstrate that the proposed Expert-CFG, with 4.2B parameters and limited expert annotations, outperforms state-of-the-art models with 13B parameters. The results demonstrate the feasibility of deploying such a system in resource-limited settings for clinical use.

cs.CV↗

arXiv2025

MST-Distill: Mixture of Specialized Teachers for Cross-Modal Knowledge Distillation

Knowledge distillation as an efficient knowledge transfer technique, has achieved remarkable success in unimodal scenarios. However, in cross-modal settings, conventional distillation methods encounter significant challenges due to data and statistical heterogeneities, failing to leverage the complementary prior knowledge embedded in cross-modal teacher models. This paper empirically reveals two critical issues in existing approaches: distillation path selection and knowledge drift. To address these limitations, we propose MST-Distill, a novel cross-modal knowledge distillation framework featuring a mixture of specialized teachers. Our approach employs a diverse ensemble of teacher models across both cross-modal and multimodal configurations, integrated with an instance-level routing network that facilitates adaptive and dynamic distillation. This architecture effectively transcends the constraints of traditional methods that rely on monotonous and static teacher models. Additionally, we introduce a plug-in masking module, independently trained to suppress modality-specific discrepancies and reconstruct teacher representations, thereby mitigating knowledge drift and enhancing transfer effectiveness. Extensive experiments across five diverse multimodal datasets, spanning visual, audio, and text, demonstrate that our method significantly outperforms existing state-of-the-art knowledge distillation methods in cross-modal distillation tasks. The source code is available at https://github.com/Gray-OREO/MST-Distill.

cs.CV↗

arXiv2025

Autonomous Collaborative Scheduling of Time-dependent UAVs, Workers and Vehicles for Crowdsensing in Disaster Response

Natural disasters have caused significant losses to human society, and the timely and efficient acquisition of post-disaster environmental information is crucial for the effective implementation of rescue operations. Due to the complexity of post-disaster environments, existing sensing technologies face challenges such as weak environmental adaptability, insufficient specialized sensing capabilities, and limited practicality of sensing solutions. This paper explores the heterogeneous multi-agent online autonomous collaborative scheduling algorithm HoAs-PALN, aimed at achieving efficient collection of post-disaster environmental information. HoAs-PALN is realized through adaptive dimensionality reduction in the matching process and local Nash equilibrium game, facilitating autonomous collaboration among time-dependent UAVs, workers and vehicles to enhance sensing scheduling. (1) In terms of adaptive dimensionality reduction during the matching process, HoAs-PALN significantly reduces scheduling decision time by transforming a five-dimensional matching process into two categories of three-dimensional matching processes; (2) Regarding the local Nash equilibrium game, HoAs-PALN combines the softmax function to optimize behavior selection probabilities and introduces a local Nash equilibrium determination mechanism to ensure scheduling decision performance. Finally, we conducted detailed experiments based on extensive real-world and simulated data. Compared with the baselines (GREEDY, K-WTA, MADL and MARL), HoAs-PALN improves task completion rates by 64.12%, 46.48%, 16.55%, and 14.03% on average, respectively, while each online scheduling decision takes less than 10 seconds, demonstrating its effectiveness in dynamic post-disaster environments.

cs.MA↗

arXiv2025

Text to Image Generation and Editing: A Survey

Text-to-image generation (T2I) refers to the text-guided generation of high-quality images. In the past few years, T2I has attracted widespread attention and numerous works have emerged. In this survey, we comprehensively review 141 works conducted from 2021 to 2024. First, we introduce four foundation model architectures of T2I (autoregression, non-autoregression, GAN and diffusion) and the commonly used key technologies (autoencoder, attention and classifier-free guidance). Secondly, we systematically compare the methods of these studies in two directions, T2I generation and T2I editing, including the encoders and the key technologies they use. In addition, we also compare the performance of these researches side by side in terms of datasets, evaluation metrics, training resources, and inference speed. In addition to the four foundation models, we survey other works on T2I, such as energy-based models and recent Mamba and multimodality. We also investigate the potential social impact of T2I and provide some solutions. Finally, we propose unique insights of improving the performance of T2I models and possible future development directions. In summary, this survey is the first systematic and comprehensive overview of T2I, aiming to provide a valuable guide for future researchers and stimulate continued progress in this field.

cs.CV↗

arXiv2025

Enhancing the coherence time of a neutral atom by an optical quartic trap

The coherence time of an optically trapped neutral atom is a crucial parameter for quantum technologies. We found that optical dipole traps with higher-order spatial forms inherently offer lower decoherence rates compared to those with lower-order spatial forms. We formulated the decoherence rate caused by the variance of the differential energy shift and photon jumping rate. Then, we constructed blue-detuned harmonic and quartic optical dipole traps, and experimentally investigated the coherence time of a trapped single cesium atom. The experimental results qualitatively verified our theory. Our approach provides a novel method to enhance the coherence time of optically trapped neutral atoms.

quant-ph↗

arXiv2025

Cavity-Enhanced Rydberg Atomic Superheterodyne Receiver

High-sensitivity measurements of the microwave electric field are important in applications of communication and metrology. \replaced{The sensitivity of traditional Rydberg superheterodyne receivers in free space is effectively determined by the signal-to-noise ratio (SNR), which is often considered equivalent to sensitivity in practical sensing applications.}{The sensitivity of the traditional Rydberg superheterodyne receivers in free space is limited by signal-to-noise contrast.} In this work, we demonstrate a cavity-enhanced receiver, where an optical cavity significantly amplifies the interaction between the probe light and cesium atoms, which substantially improves the signal-to-noise ratio via enhancing the expansion coefficient $ \kappa $. \added{Here, $\kappa$ is the edge slope of the single peak obtained by fitting the double-peak EIT-AT spectrum, characterizing the response of the probe light to the frequency detuning of the coupling laser.}The sensitivity is thus boosted by a factor of approximately 19 dB. This study highlights the pivotal role of optical cavities in advancing Rydberg-based detection systems, offering a promising approach for high-sensitivity microwave electric field measurements.

quant-ph↗

arXiv2025

High-precision measurement of microwave electric field by cavity-enhanced critical behavior in a many-body Rydberg atomic system

It has been demonstrated that the Rydberg criticality in a many-body atomic system can enhance the measurement sensitivity of the microwave electric field by increasing the Fisher information. In our previous work, we proposed and experimentally verified that the Fisher information near the critical point can be increased by more than two orders of magnitude with the Rydberg atoms coupled with an optical cavity compared with that in free space. Here we demonstrate the precision measurement of the microwave electric field by cavity-enhanced critical behavior. We show that the equivalent measurement sensitivity of the microwave electric field can be enhanced by an order of magnitude compared with that in free space. The obtained sensitivity can be enhanced to 2.6 nV/cm/Hz$^{1/2}$.

quant-ph↗

arXiv2025

A cavity QED system with defect-free single-atom array strongly coupled to an optical cavity

We experimentally realize a new cavity quantum electrodynamics (QED) platform with defect-free single-atom array strongly coupled to an optical cavity. The defect-free single-atom array is obtained by rearranging a probabilistically loaded one-dimensional (1D) optical tweezer array with dimensions of $1 \times 40$. The atom array is enclosed with two cavity mirrors, which compose a miniature optical Fabry-P{\'e}rot cavity with cavity length of 1.15 mm. By precisely controlling the position of the atom array, we demonstrate uniform and strong coupling of all atoms in the array with the optical cavity. The average coupling strength between the single atom and the cavity is 2.62 MHz. The vacuum Rabi splitting spectra for single-atom arrays with atom number $N$ changing from 3 to 26 are measured. Thus, the collective enhancement of the coupling strength with ${\sqrt N}$-dependence for multiple atoms is validated at the single atom level. Our system holds significant potential for establishing the foundation of distributed quantum computing and advancing fundamental research in many-body physics.

quant-ph↗

arXiv2024

Training Verification-Friendly Neural Networks via Neuron Behavior Consistency

Formal verification provides critical security assurances for neural networks, yet its practical application suffers from the long verification time. This work introduces a novel method for training verification-friendly neural networks, which are robust, easy to verify, and relatively accurate. Our method integrates neuron behavior consistency into the training process, making neuron activation states remain consistent across different inputs within a local neighborhood. This reduces the number of unstable neurons and tightens the bounds of neurons thereby enhancing the network's verifiability. We evaluated our method using the MNIST, Fashion-MNIST, and CIFAR-10 datasets with various network architectures. The experimental results demonstrate that networks trained using our method are verification-friendly across different radii and architectures, whereas other tools fail to maintain verifiability as the radius increases. Additionally, we show that our method can be combined with existing approaches to further improve the verifiability of networks.

cs.LG↗