Searcharxiv⌕ Search

arXiv subjects

Pengyu Chen

Publications and source records attributed to Pengyu Chen.

At least 19 recordsLinked to original sources
arXiv2026

MemeMind: Reference-Guided Trace Construction for Offline Context Optimization

Offline context optimization improves an agent by revising its instructions and examples while keeping the model frozen. This approach learns from rollouts on an adaptation set, but some queries produce only failed rollouts. In these cases, the optimizer sees no successful example of how the available tools can reach the correct answer. We introduce MemeMind, which uses an offline reference answer to recover this missing experience. TraceBuilder identifies the evidence required by the reference, executes text search, image retrieval, and visual grounding, and verifies the resulting tool trace before adding it to the adaptation buffer. ToolGuide then summarizes the collected traces into a shared guide and separate instructions for each tool. The reference answers and constructed traces are used only during adaptation, while inference uses the learned guides with a frozen model. We study this problem through Anime, Comic, and Game meme interpretation. These memes combine edited and ambiguous visual content, overlaid text, long tail franchise knowledge, and culture specific references. Their interpretation can require coordinated visual grounding, image retrieval, and text search, making them a demanding setting in which native rollout groups may fail together. We evaluate MemeMind on MemeX, a benchmark of 1,000 such memes annotated by experts. Across two Qwen3-VL models, two language partitions, and two independent judges, MemeMind improves over the strongest context optimization baseline by 22.0% and 21.1% on Qwen3-VL-30B-A3B, and by 8.1% and 8.0% on Qwen3-VL-235B-A22B under GPT-5 judging. Ablations and held out traces show that constructing successful tool use for failed groups provides the largest component gain and produces more effective evidence acquisition at inference time.

cs.CV↗
arXiv2026

Prox: Training-Free FFN Activation Sparsity via Approximate Intermediate-Channel Salience in LLMs

Feed-forward networks (FFNs) dominate memory traffic and computation in large language model (LLM) inference, making them a primary target for activation sparsification. However, existing training-free methods suffer substantial model-quality degradation at high sparsity due to limitations in their channel-selection strategies. We observe that the SwiGLU intermediate state provides a highly effective channel-selection signal, but obtaining it requires costly dense computation. To address this, we present \emph{Prox}, a two-stage training-free framework for sparse SwiGLU FFNs. Prox hinges on the key insight: sparse execution requires only the channel mask induced by the intermediate state, which can be constructed from the magnitude ranking of its entries rather than their exact values. Specifically, Stage 1 uses input sparsity and quantized proxy weights to construct a shared mask; Stage 2 computes the selected channels exactly, enabling sparse execution of all three projections. Across ten LLMs from six model families, Prox outperforms training-free baselines at all sparsity levels, achieves up to a $1.99\times$ end-to-end decoding speedup at 70\% FFN sparsity, and is compatible with quantization and sparse attention.

cs.LG↗
arXiv2026

MemeBench: What LVLMs Miss When Interpreting Culture-Dependent Memes

Large vision-language models have improved at describing visual content, but accurate descriptions do not ensure interpretation when meaning depends on knowledge beyond the pixels. Memes expose this gap because they rely on cultural entities, background knowledge, and community conventions. Most meme benchmarks reduce interpretation to labels or holistic scores, obscuring where an explanation breaks down. We introduce MemeBench, a diagnostic benchmark of 1,253 Chinese and English memes with human-written references and quality-controlled VIKR annotations, centered on anime, comics, games, and adjacent online subcultures. Its VIKR schema decomposes explanations into Visual clues, Identity links, Knowledge units, and Reasoning mechanisms. Across 26 LVLMs, every model covers visible content more reliably than the knowledge needed to interpret it, and even the strongest retains a 22.6% Visual-Knowledge gap. To test whether this diagnosis can guide improvement, we introduce KAR, an entity-guided retrieval baseline built on CultureBase. Across four controlled models, KAR raises VIKR Success by 3.6-7.4% and, compared with generic retrieval, repairs more answers and breaks fewer. Yet both retrieval conditions improve Identity and Knowledge while reducing Visual coverage in every comparison. MemeBench reveals whether an interpretation succeeds, what is missing, and whether targeted evidence fills the diagnosed gap.

cs.AI↗
arXiv2026

Demystifying Deep Learning Compiler Frontend Bugs: An LLM-Aided Empirical Study

Deep learning compilers (DLCs) are designed to translate deep learning programs into optimized, hardware-specific code. Typically, DLC frontends translate programs into graph-based intermediate representations (IRs) to enable optimizations. Defects introduced during this stage (termed \emph{fBug}s) are severe yet understudied, as prior work predominantly focuses on low-level APIs and operators or treats DLCs as monolithic entities. To bridge this gap, we conduct the first systematic empirical study of \emph{fBug}s in TorchDynamo, the default DLC frontend for PyTorch 2, the most popular DL framework. Leveraging a domain-knowledge-enhanced LLM-aided methodology, we analyze 123 \emph{fBug}s and construct a taxonomy comprising 7 root cause categories and 15 subcategories. Our findings provide actionable insights for DLC development and testing. Furthermore, we leverage the LLM to generate targeted, root cause-aware test cases to detect new bugs. We uncovered 23 previously unknown \emph{fBug}s in recent releases (15 confirmed) across eight (sub)categories, demonstrating the efficacy of our methodology in testing and hardening DLC frontends.

cs.PL↗
arXiv2026

saCI: An R Package for Stochastic Approximation Confidence Intervals for Correlation Coefficients

This paper presents saCI, an R package that implements the stochastic approximation method for constructing nonparametric confidence intervals for Pearson's correlation coefficient. The package is based on the algorithm proposed by Garthwaite (1996) and further developed by Xiong & Xu (2016). The implementation provides both the stochastic approximation (SA) method and the bootstrap BCa method for comparison, along with an interactive Shiny application for exploratory analysis. The package has been successfully published on CRAN, demonstrating its compliance with R package standards and reproducibility.

stat.ME↗
arXiv2026

FGRPO: Federated GRPO with Adaptive Aggregation on Non-IID Data

Recent advances in language models have established reinforcement learning as the primary paradigm for eliciting self-correction and long-chain reasoning. While group relative policy optimization (GRPO) offers superior scalability by eliminating the critic network, deploying it on a central infrastructure entails collecting a large volume of data from distributed owners, which poses significant privacy risks. To address these concerns, we introduce federated GRPO (FGRPO), a framework designed to decentralize the fine-tuning of reasoning models across heterogeneous data owners. To effectively mitigate the instability caused by divergent reward scales across heterogeneous tasks, FGRPO incorporates an adaptive aggregation mechanism based on relative performance gain. By characterizing each client's improvement relative to its personalized historical baseline, the framework dynamically prioritizes effective learning trajectories regardless of local task difficulty. FGRPO ensures robust convergence on non-IID data while preserving data privacy.

cs.LG↗
arXiv2026

Scaling Multi-Hop Training Data via Graph-Constrained Path Selection

Endowing large language models with compositional reasoning over specialized documents requires multi-hop training data at scale, where such data rarely exists outside of curated benchmarks built on structured sources. To construct it directly from plain, unannotated text, existing methods ask a single teacher model to jointly discover an evidence path through a document and verbalize it as a question-answer pair. However, these methods degrade sharply when documents are structured around repetitive templates and densely cross-referencing clauses, conditions that characterize most real-world specialized corpora. In this work, we decouple the two operations: reasoning paths are enumerated offline over a graph of contextual keyword centroids, and the teacher is invoked only to verbalize pre-validated paths. The graph enforces five geometric admissibility constraints, for which we provide Gram-matrix arguments establishing that local similarity bounds alone admit endpoint drift up to ${\sim}91^{\circ}$, and that an upper similarity bound is necessary to exit dense embedding cliques formed by boilerplate text. A matched-size ablation isolates the mechanism: at equal training scale, constrained and unconstrained chains yield indistinguishable downstream performance, and the gain at full scale comes from a 4.4$\times$ expansion of the usable corpus rather than from higher per-chain quality -- reframing the role of graph constraints, in this setting, as raising teacher synthesizability rather than improving chain content. Fine-tuning Qwen3-32B on 80K examples constructed from the CUAD legal contract corpus improves closed-book Token F1 from 21.66% to 38.58%. We have released our codes at https://github.com/hkgai-official/GCSCS.

cs.CL↗
arXiv2026

STREAM: A Data-Centric Framework for Mining High-Value Task-Oriented Dialogues from Streaming Media

Large language models for vertical domains are bottlenecked by the scarcity of complex, domain-specific task-oriented dialogues. Existing data acquisition pipelines face a persistent trilemma: expert annotation is expensive, real-world service conversations are constrained by privacy and commercial restrictions, and static corpora quickly become temporally stale. We propose Stream, a data-centric framework that leverages publicly available streaming media (live streams and short videos) to synthesize high-value service dialogues at scale. Stream mines authentic interaction signals from noisy streams and synthesizes conversations by integrating role-grounded persona construction with Conversational Blueprint construction; it further adopts retrieval-augmented generation (RAG) to support knowledge-aware responses. Based on Stream, we release StreamDial, a large-scale multi-domain dataset covering Automotive, Restaurant, and Hotel. StreamDial contains 87,498 dialogue sessions and 1,497,320 turns in total, with an average of 17.11 turns per session and a comparable scale across domains. Each session is organized as a structured quadruplet $\langle P_u, P_a, B, H \rangle$ that pairs dialogue history with explicit user/agent personas and a Conversational Blueprint, capturing realistic service behaviors such as requirement mining, constraint conflicts, negotiation, and recovery. Evaluations with automatic judges and downstream tasks show that StreamDial improves intrinsic dialogue quality over strong baselines, and models trained with StreamDial improve Dialogue State Tracking across backbones; we further report a completed human-evaluation set and encouraging multilingual transfer on Qwen3-8B under a controlled training budget. The data is released in https://github.com/hitxueliang/DialogDataSetBySTREAM.

cs.CL↗
arXiv2026

FG-TreeSeg: Flow-Guided Tree Crown Segmentation without Instance Annotations

Individual tree crown segmentation is an important task in remote sensing for forest biomass estimation and ecological monitoring. However, accurate delineation in dense, overlapping canopies remains a bottleneck. While supervised deep learning methods suffer from high annotation costs and limited generalization, emerging foundation models (e.g., Segment Anything Model) often lack domain knowledge, leading to under-segmentation in dense clusters. To bridge this gap, we propose FG-TreeSeg, a training-free framework for tree crown instance segmentation that transfers flow-based delineation from biomedical imaging to remote sensing. By modeling tree crowns as star-convex objects within a topological flow field using Cellpose-SAM, the FG-TreeSeg framework forces the separation of touching tree crown instances based on vector convergence. Experiments on the NEON and BAMFOREST datasets and visual inspection demonstrate that our framework generalizes robustly across diverse sensor types and canopy densities, which can offer a training-free solution for tree crown instance segmentation and labels generation.

cs.CV↗
arXiv2026

Model Forensics in AI-Native Wireless Networks: Taxonomy, Applications, and Case Study

As artificial intelligence (AI) is increasingly embedded in wireless networks, models are becoming core components that influence signal processing, resource scheduling and network control. However, model anomalies, tampering and malicious functions also introduce new security risks. In this article, we focus on model forensics in AI-native wireless networks. Specifically, we first discuss key problems including model authenticity verification, malicious function identification and accountability tracing, and summarize the main categories of model forensics. We then explain the role of model forensics in AI-native wireless networks and review representative application scenarios. In the case study, we use RF fingerprinting as an example and present two concrete workflows based on watermark authentication and backdoor detection, illustrating how provenance authentication and malicious behavior identification can be implemented in practice. The results show that model forensics can provide important support for anomaly assessment, provenance tracing and trustworthy operation in AI-native wireless networks. Finally, we outline several promising directions for future research in this emerging area.

cs.CR↗
arXiv2026

Intelligent bear deterrence system based on computer vision: Reducing human-bear conflicts in remote areas

Human-bear conflicts on the Tibetan Plateau threaten both local livelihoods and the conservation of Tibetan brown bears (Ursus arctos pruinosus). To address this challenge, we developed a low-power, network-independent deterrence system that combines computer vision with Internet of Things (IoT) hardware. The system integrates a YOLOv5-MobileNet detection model deployed on a low-power edge artificial intelligence (AI) board with a solar-powered bear spray device. We compiled a data set of 1,243 wildlife images (including 795 bears with 100 infrared captures for nighttime detection, plus other common objects and animals such as mastiffs, yaks, humans, and vehicles), from which 80% were used for training and 20% for validation. Validation showed robust performance (mean average precision = 91.4%, recall = 93.6%). In 100 controlled activation tests involving simulated approaches by bears, humans, and other animals, the spray deployed within 0.2 seconds of detection with 97.2% accuracy, confirming timely and reliable responses. A 30-day field trial in Zadoi County, Qinghai Province, China, recorded 3 successful deterrence events without false activations. By using energy-efficient components and ensuring continuous and stable system operation, this solution provides a practical, sustainable, and scalable approach to mitigating human-bear conflicts, effectively enhancing human safety and bear conservation in remote areas without network or grid coverage.

cs.CV↗
arXiv2026

Cross-Modal Urban Sensing: Evaluating Sound-Vision Alignment Across Street-Level and Aerial Imagery

Environmental soundscapes convey substantial ecological and social information regarding urban environments; however, their potential remains largely untapped in large-scale geographic analysis. In this study, we investigate the extent to which urban sounds correspond with visual scenes by comparing various visual representation strategies in capturing acoustic semantics. We employ a multimodal approach that integrates geo-referenced sound recordings with both street-level and remote sensing imagery across three major global cities: London, New York, and Tokyo. Utilizing the AST model for audio, along with CLIP and RemoteCLIP for imagery, as well as CLIPSeg and Seg-Earth OV for semantic segmentation, we extract embeddings and class-level features to evaluate cross-modal similarity. The results indicate that street view embeddings demonstrate stronger alignment with environmental sounds compared to segmentation outputs, whereas remote sensing segmentation is more effective in interpreting ecological categories through a Biophony--Geophony--Anthrophony (BGA) framework. These findings imply that embedding-based models offer superior semantic alignment, while segmentation-based methods provide interpretable links between visual structure and acoustic ecology. This work advances the burgeoning field of multimodal urban sensing by offering novel perspectives for incorporating sound into geospatial analysis.

cs.CV↗
arXiv2026

SDD-YOLO: A Small-Target Detection Framework for Ground-to-Air Anti-UAV Surveillance with Edge-Efficient Deployment

Detecting small unmanned aerial vehicles (UAVs) from a ground-to-air (G2A) perspective presents significant challenges, including extremely low pixel occupancy, cluttered aerial backgrounds, and strict real-time constraints. Existing YOLO-based detectors are primarily optimized for general object detection and often lack adequate feature resolution for sub-pixel targets, while introducing complexities during deployment. In this paper, we propose SDD-YOLO, a small-target detection framework tailored for G2A anti-UAV surveillance. To capture fine-grained spatial details critical for micro-targets, SDD-YOLO introduces a P2 high-resolution detection head operating at 4 times downsampling. Furthermore, we integrate the recent architectural advancements from YOLO26, including a DFL-free, NMS-free architecture for streamlined inference, and the MuSGD hybrid training strategy with ProgLoss and STAL, which substantially mitigates gradient oscillation on sparse small-target signals. To support our evaluation, we construct DroneSOD-30K, a large-scale G2A dataset comprising approximately 30,000 annotated images covering diverse meteorological conditions. Experiments demonstrate that SDD-YOLO-n achieves a mAP@0.5 of 86.0% on DroneSOD-30K, surpassing the YOLOv5n baseline by 7.8 percentage points. Extensive inference analysis shows our model attains 226 FPS on an NVIDIA RTX 5090 and 35 FPS on an Intel Xeon CPU, demonstrating exceptional efficiency for future edge deployment.

cs.CV↗
arXiv2026

Cross-Problem Solving for Network Optimization: Is Problem-Aware Learning the Key?

As intelligent network services continue to diversify, ensuring efficient and adaptive resource allocation in edge networks has become increasingly critical. Yet the wide functional variations across services often give rise to new and unforeseen optimization problems, rendering traditional manual modeling and solver design both time-consuming and inflexible. This limitation reveals a key gap between current methods and human solving - the inability to recognize and understand problem characteristics. It raises the question of whether problem-aware learning can bridge this gap and support effective cross-problem generalization. To answer this question, we propose a problem-aware diffusion (PAD) model, which leverages a problem-aware learning framework to enable cross-problem generalization. By explicitly encoding the mathematical formulations of optimization problems into token-level embeddings, PAD empowers the model to understand and adapt to problem structures. Extensive experiments across ten representative network optimization problems show that PAD generalizes well to unseen problems while avoiding the inefficiency of building new solvers from scratch, yet still delivering competitive solution quality. Meanwhile, an auxiliary constraint-aware module is designed to enforce solution validity further. The experiments indicate that problem-aware learning opens a promising direction toward general-purpose solvers for intelligent network operation and resource management. Our code is open source at https://github.com/qiyu3816/PAD.

cs.NI↗
arXiv2026

TextFlux: An OCR-Free DiT Model for High-Fidelity Multilingual Scene Text Synthesis

Diffusion-based scene text synthesis has progressed rapidly, yet existing methods commonly rely on additional visual conditioning modules and require large-scale annotated data to support multilingual generation. In this work, we revisit the necessity of complex auxiliary modules and further explore an approach that simultaneously ensures glyph accuracy and achieves high-fidelity scene integration, by leveraging diffusion models' inherent capabilities for contextual reasoning. To this end, we introduce TextFlux, a DiT-based framework that enables multilingual scene text synthesis. The advantages of TextFlux can be summarized as follows: (1) OCR-free model architecture. TextFlux eliminates the need for OCR encoders (additional visual conditioning modules) that are specifically used to extract visual text-related features. (2) Strong multilingual scalability. TextFlux is effective in low-resource multilingual settings, and achieves strong performance in newly added languages with fewer than 1,000 samples. (3) Streamlined training setup. TextFlux is trained with only 1% of the training data required by competing methods. (4) Controllable multi-line text generation. TextFlux offers flexible multi-line synthesis with precise line-level control, outperforming methods restricted to single-line or rigid layouts. Extensive experiments and visualizations demonstrate that TextFlux outperforms previous methods in both qualitative and quantitative evaluations.

cs.CV↗
arXiv2026

Mode Conversion of Hyperbolic Phonon Polaritons in van der Waals terraces

Electromagnetic hyperbolicity has driven key functionalities in nanophotonics, including super-resolution imaging, efficient energy control, and extreme light manipulation. Central to these advances are hyperbolic polaritons - nanometer-scale light-matter waves - spanning multiple energy-momentum dispersion orders with distinct mode profiles and incrementally high optical momenta. In this work, we report the mode conversion of hyperbolic polaritons across different dispersion orders by breaking the structure symmetry in engineered step-shape van der Waals (vdW) terraces. The mode conversion from the fundamental to high-order hyperbolic polaritons is imaged using scattering-type scanning near-field optical microscopy (s-SNOM) on both hexagonal boron nitride (hBN) and alpha-phase molybdenum trioxide (alpha-MoO3) vdW terraces. Our s-SNOM data, augmented with electromagnetic simulations, further demonstrate the alteration of polariton mode conversion by varying the step size of vdW terraces. The mode conversion reported here offers a practical approach toward integrating previously independent different-order hyperbolic polaritons with ultra-high momenta, paving the way for promising applications in nano-optical circuits, sensing, computation, information processing, and super-resolution imaging.

physics.optics↗
arXiv2025

CoDrone: Autonomous Drone Navigation Assisted by Edge and Cloud Foundation Models

Autonomous navigation for Unmanned Aerial Vehicles faces key challenges from limited onboard computational resources, which restrict deployed deep neural networks to shallow architectures incapable of handling complex environments. Offloading tasks to remote edge servers introduces high latency, creating an inherent trade-off in system design. To address these limitations, we propose CoDrone - the first cloud-edge-end collaborative computing framework integrating foundation models into autonomous UAV cruising scenarios - effectively leveraging foundation models to enhance performance of resource-constrained unmanned aerial vehicle platforms. To reduce onboard computation and data transmission overhead, CoDrone employs grayscale imagery for the navigation model. When enhanced environmental perception is required, CoDrone leverages the edge-assisted foundation model Depth Anything V2 for depth estimation and introduces a novel one-dimensional occupancy grid-based navigation method - enabling fine-grained scene understanding while advancing efficiency and representational simplicity of autonomous navigation. A key component of CoDrone is a Deep Reinforcement Learning-based neural scheduler that seamlessly integrates depth estimation with autonomous navigation decisions, enabling real-time adaptation to dynamic environments. Furthermore, the framework introduces a UAV-specific vision language interaction module incorporating domain-tailored low-level flight primitives to enable effective interaction between the cloud foundation model and the UAV. The introduction of VLM enhances open-set reasoning capabilities in complex unseen scenarios. Experimental results show CoDrone outperforms baseline methods under varying flight speeds and network conditions, achieving a 40% increase in average flight distance and a 5% improvement in average Quality of Navigation.

cs.RO↗
arXiv2025

Subcellular Metabolic Tracking Using Fluorescent Nanodiamonds Relaxometry

Fluorescent nanodiamonds (FNDs) relaxometry holds promising future for advancement of high spatiotemporal resolution metabolic imaging technology. In this study, we demonstrate a simultaneous integration of spatial position tracking with FND relaxometry to characterize the temporal dynamics of metabolic processes, thereby enhancing the capability to monitor cellular activities over time. To enable targeted metabolic probing in living cells, FNDs were surface-functionalized to achieve specific localization within key organelles, including the nucleus and mitochondria. This strategy not only facilitates subcellular-level metabolic monitoring but also allows for direct comparison between intra- and extranuclear microenvironments within the same living cell, offering substantial potential for elucidating the spatial and functional heterogeneity of cellular metabolism.

physics.bio-ph↗