arXiv subjects

Cheng Jiang

Publications and source records attributed to Cheng Jiang.

At least 19 recordsLinked to original sources

arXiv2026

SE-GoS: Self-Evolving Graph-of-Skills for Skill Library at Scale

Modern LLM agents increasingly rely on reusable skills, yet as skill libraries scale to thousands of entries, effective retrieval becomes a bottleneck. Graph-of-Skills (GoS) addresses this challenge by exploiting dependency-aware graph structure for scalable skill retrieval, while SkillDAG further demonstrates that skill graphs can accumulate execution-backed structure online. However, these approaches leave open whether historical execution traces can be systematically distilled into a better retrieval graph that generalizes to unseen tasks. We present Self-Evolving Graph-of-Skills (SE-GoS), a training-free framework that evolves an existing GoS graph from execution traces while preserving the original retrieval pipeline. SE-GoS performs three complementary updates: topology evolution that discovers and prunes skill relationships from execution evidence, edge-weight evolution that reinforces retrieval-relevant relationships based on historical effectiveness, and description evolution that optimizes retrieval-facing skill descriptions using execution feedback. Across three LLMs on SkillsBench, SE-GoS consistently improves task reward while reducing input tokens relative to full skill loading, with gains varying across model families. In a representative setting, one evolution round improves reward from 52.4\% to 59.4\% while reducing input tokens by approximately one-third relative to full skill loading, and the resulting graph transfers to a disjoint held-out split with a 5.4-point improvement over the static GoS baseline. These results show that skill graphs can be improved from execution experience without model training, changes to the retrieval algorithm, or modifications to skill content, turning a static retrieval graph into an evolving retrieval infrastructure.

cs.AI↗

arXiv2026

True Work-Efficiency in Parallel Derandomization

A longstanding limitation of known techniques for parallel derandomization was that they incurred at least polylogarithmic overhead in work. For instance, for fundamental and frequently used problems such as maximal independent set, maximal matching, and $(\Delta+1)$-coloring, where $\Delta$ denotes the maximum degree of the graph, the best-known deterministic parallel algorithms with polylogarithmic depth used $\Omega((m+n)\mathrm{poly}(\log n))$ work on $n$-vertex, $m$-edge graphs; see, e.g., Luby [FOCS '88]. Consequently, at least $\mathrm{poly}(\log n)$ processors were needed for these algorithms to outperform straightforward single-processor algorithms. Recently, Ghaffari and Grunau [FOCS '25] introduced a new parallel derandomization method that substantially reduced the overhead from $\mathrm{poly}(\log n)$ to $\mathrm{poly}(\log\log n)$, achieving work bounds of $O((m+n)\mathrm{poly}(\log\log n))$. In this paper, we settle this line of research by obtaining linear work bounds of $O(m+n)$, thereby achieving truly work-efficient parallel derandomization.

cs.DS↗

arXiv2026

Agentic Hybrid RAG for Evidence-Grounded Muon Collider Analysis

Muon collider research spans accelerator physics, detector instrumentation, and high-energy phenomenology, with relevant evidence scattered across a rapidly expanding and heterogeneous body of scientific literature. As high-energy physics (HEP) increasingly explores agent-assisted analysis workflows, efficiently locating, integrating, and verifying scientific evidence becomes an essential capability. While retrieval-augmented generation (RAG) offers a promising framework for scientific question answering, integrating agentic reasoning without compromising retrieval precision remains a key challenge. In this work, we present agentic hybrid RAG, an evidence-grounded RAG framework for muon collider research. The framework combines a hybrid retriever, integrating sparse lexical and dense semantic retrieval, with an agentic reasoning module for query decomposition, evidence expansion, and grounded answer generation. To enable systematic evaluation, we construct the first benchmark for retrieval-augmented scientific question answering in the muon collider domain, comprising a curated literature corpus together with dedicated retrieval and answer-generation benchmarks covering major detector and physics research topics. Extensive evaluation shows that hybrid retrieval provides the strongest retrieval backbone, while agentic reasoning is most effective for controlled evidence expansion and answer synthesis. Built on this principle, agentic hybrid RAG consistently outperforms representative retrieval and RAG baselines in retrieval effectiveness, answer quality, evidence coverage, and factual grounding. Together, the benchmark and framework provide a foundation for evidence-grounded scientific question answering and future HEP analysis agents operating over large-scale scientific literature.

hep-ex↗

arXiv2026

CaloTrilogy: Toward a Breakthrough in One-Step, End-to-End, Physics-Guided Shower Generation for Modern Calorimeters

High-precision calorimeter simulation at current and future colliders imposes rapidly growing computational demands, motivating the development of machine-learning surrogates for traditional Monte Carlo tools such as Geant4. Flow matching and diffusion-based generative models have become leading approaches for high-dimensional fast simulation because of their sample quality, but typically require ${\cal O}(100)$ function evaluations at inference and often rely on auxiliary networks to constrain global observables, compromising streamlined end-to-end generation. We introduce a unified framework that improves the balance between speed, shower quality, and physics fidelity. The method combines: (i) an average velocity field integrator that enables sampling in one or a few evaluations; (ii) a learned generative prior in shower space, constructed from data rather than random noise; and (iii) physics-guided loss terms that impose inductive biases on key observables during training. These elements are training time regularizers, preserving end-to-end inference with no additional cost. With only one or a few evaluation steps, the model achieves shower quality competitive with state-of-the-art flow and diffusion approaches, tested on several public high granularity calorimeter datasets. The results demonstrate inter-layer shower structure consistent with the underlying physics, providing a strong candidate for future fast simulation workflows.

hep-ex↗

arXiv2026

Subspace-Guided Semantic and Topological Invariant Registration for Annotation-Free Ultrasound Plane Quality Control

Reliable quality control (QC) of ultrasound images is essential for both real-time acquisition guidance and retrospective clinical audit, yet existing approaches rely heavily on per-plane annotations, or employ pseudo-labeling prone to systematic bias under spatial deformations inherent in clinical acquisition. We present STRIQ, a registration-driven framework that recasts annotation-free US plane quality control as a subspace-guided consistency measurement problem. Specifically, STRIQ introduces a Latent Registration Aligner (LRA) to establish hierarchical feature space correspondences between query images and variance-driven anchors, which are autonomously distilled from unlabeled data via a variance spectrum criterion to serve as structurally stable prototypes. To further disambiguate anatomical planes and mitigate negative knowledge transfer, we propose an Orthogonal Knowledge Subspace (OKS) module. The OKS decomposes plane-specific representations into mutually orthogonal subspaces, enabling fine-grained expert collaboration while preventing inter-plane interference, ensuring that the quality metric is grounded in principled subspace proximity. Extensive experiments on the in-house US4QA and public CAMUS datasets demonstrate that STRIQ achieves state-of-the-art correlation with clinical quality scores, establishing a new paradigm for annotation-free, real-time reliable ultrasound quality control. Our code is available at https://github.com/zhcz328/STRIQ.

cs.CV↗

arXiv2026

Turnstile Streaming Algorithms Might (Still) as Well Be Linear Sketches, for Polynomial-Length Streams

A fundamental question in streaming complexity is whether every space-efficient turnstile algorithm is implicitly a linear sketch. The landmark work of Li, Nguyen, and Woodruff [LNW14] established an equivalence between the two, but their reduction requires a stream length that is at least doubly exponential in the dimension $n$. In the opposite direction, results by Kallaugher and Price [KP20] demonstrate a separation for streams of linear length, showing that the equivalence does not hold in general. The most natural and practically relevant regime -- polynomial-length streams -- has therefore remained open. We show that polynomial-length turnstile algorithms admit linear-sketch simulations. More precisely, if a turnstile algorithm uses $S$ bits of space and succeeds on all streams of length $\mathrm{poly}(D, n)$, then on final vectors $x$ with $\|x\|_2 \le D$, its output can be recovered from $O(S)$ linear measurements of $x$, using $O(S \log S)$ bits overall. For smooth problems under appropriate input distributions, a mollified version of the reduction yields a bounded-entry sketch with $O(S / \log D)$ measurements and optimal $O(S)$ total space. Our results extend to strict turnstile streams and non-uniform Read-Once Branching Programs (ROBPs). Our proof departs from prior transition-graph based machinery, relying instead on a Fourier-analytic framework and tools from additive combinatorics to extract discrete linear measurements. Our analysis shows that any $S$-bit algorithm can only be sensitive to a low-dimensional lattice of heavy Fourier frequencies, which we then use to construct the rows of the sketching matrix. Consequently, we obtain new lower bounds for polynomial-length streams via existing real sketching and communication lower bounds.

cs.DS↗

arXiv2026

Rapidly deploying on-device eye tracking by distilling visual foundation models

Eye tracking (ET) plays a critical role in augmented and virtual reality applications. However, rapidly deploying high-accuracy, on-device gaze estimation for new products remains challenging because hardware configurations (e.g., camera placement, camera pose, and illumination) often change across device generations. Visual foundation models (VFMs) excel on natural-image benchmarks and offer a promising path to rapid training and deployment; yet, we find that off-the-shelf VFMs still struggle to reach high accuracy on specialized near-eye infrared images. To close this gap, we introduce DistillGaze, a framework that distills a VFM using labeled synthetic data and unlabeled real data for rapid, high-accuracy on-device gaze estimation. DistillGaze proceeds in two stages. First, we adapt a VFM into a domain-specialized teacher using synthetic gaze labels and unlabeled real images. Synthetic data provide scalable, high-quality gaze supervision, while unlabeled real data bridges the synthetic-to-real domain gap. Second, we train an on-device student from both teacher guidance and self-training. Evaluated on a large-scale crowd-sourced dataset spanning more than 2,000 participants, DistillGaze reduces median gaze error by 58.6% relative to synthetic-only baselines while maintaining a lightweight 256K-parameter model suitable for real-time on-device deployment. More broadly, DistillGaze offers an efficient path to training and deploying ET models that adapt to hardware changes, and a recipe for combining synthetic supervision with unlabeled real data in on-device regression tasks.

cs.CV↗

arXiv2026

Incorporating contextual information into KGWAS for interpretable GWAS discovery

Genome-Wide Association Studies (GWAS) identify associations between genetic variants and disease; however, moving beyond associations to causal mechanisms is critical for therapeutic target prioritization. The recently proposed Knowledge Graph GWAS (KGWAS) framework addresses this challenge by linking genetic variants to downstream gene-gene interactions via a knowledge graph (KG), thereby improving detection power and providing mechanistic insights. However, the original KGWAS implementation relies on a large general-purpose KG, which can introduce spurious correlations. We hypothesize that cell-type specific KGs from disease-relevant cell types will better support disease mechanism discovery. Here, we show that the general-purpose KG in KGWAS can be substantially pruned with no loss of statistical power on downstream tasks, and that performance further improves by incorporating gene-gene relationships derived from perturb-seq data. Importantly, using a sparse, context-specific KG from direct perturb-seq evidence yields more consistent and biologically robust disease-critical networks.

cs.LG↗

arXiv2025

Scale What Counts, Mask What Matters: Evaluating Foundation Models for Zero-Shot Cross-Domain Wi-Fi Sensing

While Wi-Fi sensing offers a compelling, privacy-preserving alternative to cameras, its practical utility has been fundamentally undermined by a lack of robustness across domains. Models trained in one setup fail to generalize to new environments, hardware, or users, a critical "domain shift" problem exacerbated by modest, fragmented public datasets. We shift from this limited paradigm and apply a foundation model approach, leveraging Masked Autoencoding (MAE) style pretraining on the largest and most heterogeneous Wi-Fi CSI datasets collection assembled to date. Our study pretrains and evaluates models on over 1.3 million samples extracted from 14 datasets, collected using 4 distinct devices across the 2.4/5/6 GHz bands and bandwidths from 20 to 160 MHz. Our large-scale evaluation is the first to systematically disentangle the impacts of data diversity versus model capacity on cross-domain performance. The results establish scaling trends on Wi-Fi CSI sensing. First, our experiments show log-linear improvements in unseen domain performance as the amount of pretraining data increases, suggesting that data scale and diversity are key to domain generalization. Second, based on the current data volume, larger model can only provide marginal gains for cross-domain performance, indicating that data, rather than model capacity, is the current bottleneck for Wi-Fi sensing generalization. Finally, we conduct a series of cross-domain evaluations on human activity recognition, human gesture recognition and user identification tasks. The results show that the large-scale pretraining improves cross-domain accuracy ranging from 2.2% to 15.7%, compared to the supervised learning baseline. Overall, our findings provide insightful direction for designing future Wi-Fi sensing systems that can eventually be robust enough for real-world deployment.

cs.CV↗

arXiv2025

Health system learning achieves generalist neuroimaging models

Frontier artificial intelligence (AI) models, such as OpenAI's GPT-5 and Meta's DINOv3, have advanced rapidly through training on internet-scale public data, yet such systems lack access to private clinical data. Neuroimaging, in particular, is underrepresented in the public domain due to identifiable facial features within MRI and CT scans, fundamentally restricting model performance in clinical medicine. Here, we show that frontier models underperform on neuroimaging tasks and that learning directly from uncurated data generated during routine clinical care at health systems, a paradigm we call health system learning, yields high-performance, generalist neuroimaging models. We introduce NeuroVFM, a visual foundation model trained on 5.24 million clinical MRI and CT volumes using a scalable volumetric joint-embedding predictive architecture. NeuroVFM learns comprehensive representations of brain anatomy and pathology, achieving state-of-the-art performance across multiple clinical tasks, including radiologic diagnosis and report generation. The model exhibits emergent neuroanatomic understanding and interpretable visual grounding of diagnostic findings. When paired with open-source language models through lightweight visual instruction tuning, NeuroVFM generates radiology reports that surpass frontier models in accuracy, clinical triage, and expert preference. Through clinically grounded visual understanding, NeuroVFM reduces hallucinated findings and critical errors, offering safer clinical decision support. These results establish health system learning as a paradigm for building generalist medical AI and provide a scalable framework for clinical foundation models.

cs.CV↗

arXiv2025

Learning neuroimaging models from health system-scale data

Neuroimaging is a ubiquitous tool for evaluating patients with neurological diseases. The global demand for magnetic resonance imaging (MRI) studies has risen steadily, placing significant strain on health systems, prolonging turnaround times, and intensifying physician burnout. These challenges disproportionately impact patients in low-resource and rural settings. Here, we utilized a large academic health system as a data engine to develop Prima, the first vision language model (VLM) serving as an AI foundation for neuroimaging that supports real-world, clinical MRI studies as input. Trained on over 220,000 MRI studies, Prima uses a hierarchical vision architecture that provides general and transferable MRI features. Prima was tested in a 1-year health system-wide study that included 30K MRI studies. Across 52 radiologic diagnoses from the major neurologic disorders, including neoplastic, inflammatory, infectious, and developmental lesions, Prima achieved a mean diagnostic area under the ROC curve of 92.0, outperforming other state-of-the-art general and medical AI models. Prima offers explainable differential diagnoses, worklist priority for radiologists, and clinical referral recommendations across diverse patient demographics and MRI systems. Prima demonstrates algorithmic fairness across sensitive groups and can help mitigate health system biases, such as prolonged turnaround times for low-resource populations. These findings highlight the transformative potential of health system-scale VLMs and Prima's role in advancing AI-driven healthcare.

cs.CV↗

arXiv2025

Reinforcement Learning on Pre-Training Data

The growing disparity between the exponential scaling of computational resources and the finite growth of high-quality text data now constrains conventional scaling approaches for large language models (LLMs). To address this challenge, we introduce Reinforcement Learning on Pre-Training data (RLPT), a new training-time scaling paradigm for optimizing LLMs. In contrast to prior approaches that scale training primarily through supervised learning, RLPT enables the policy to autonomously explore meaningful trajectories to learn from pre-training data and improve its capability through reinforcement learning (RL). While existing RL strategies such as reinforcement learning from human feedback (RLHF) and reinforcement learning with verifiable rewards (RLVR) rely on human annotation for reward construction, RLPT eliminates this dependency by deriving reward signals directly from pre-training data. Specifically, it adopts a next-segment reasoning objective, rewarding the policy for accurately predicting subsequent text segments conditioned on the preceding context. This formulation allows RL to be scaled on pre-training data, encouraging the exploration of richer trajectories across broader contexts and thereby fostering more generalizable reasoning skills. Extensive experiments on both general-domain and mathematical reasoning benchmarks across multiple models validate the effectiveness of RLPT. For example, when applied to Qwen3-4B-Base, RLPT yields absolute improvements of $3.0$, $5.1$, $8.1$, $6.0$, $6.6$, and $5.3$ on MMLU, MMLU-Pro, GPQA-Diamond, KOR-Bench, AIME24, and AIME25, respectively. The results further demonstrate favorable scaling behavior, suggesting strong potential for continued gains with more compute. In addition, RLPT provides a solid foundation, extending the reasoning boundaries of LLMs and enhancing RLVR performance.

cs.CL↗

arXiv2025

A Fully Automatic Framework for Intracranial Pressure Grading: Integrating Keyframe Identification, ONSD Measurement and Clinical Data

Intracranial pressure (ICP) elevation poses severe threats to cerebral function, thus necessitating monitoring for timely intervention. While lumbar puncture is the gold standard for ICP measurement, its invasiveness and associated risks drive the need for non-invasive alternatives. Optic nerve sheath diameter (ONSD) has emerged as a promising biomarker, as elevated ICP directly correlates with increased ONSD. However, current clinical practices for ONSD measurement suffer from inconsistency in manual operation, subjectivity in optimal view selection, and variability in thresholding, limiting their reliability. To address these challenges, we introduce a fully automatic two-stage framework for ICP grading, integrating keyframe identification, ONSD measurement and clinical data. Specifically, the fundus ultrasound video processing stage performs frame-level anatomical segmentation, rule-based keyframe identification guided by an international consensus statement, and precise ONSD measurement. The intracranial pressure grading stage then fuses ONSD metrics with clinical features to enable the prediction of ICP grades, thereby demonstrating an innovative blend of interpretable ultrasound analysis and multi-source data integration for objective clinical evaluation. Experimental results demonstrate that our method achieves a validation accuracy of $0.845 \pm 0.071$ (with standard deviation from five-fold cross-validation) and an independent test accuracy of 0.786, significantly outperforming conventional threshold-based method ($0.637 \pm 0.111$ validation accuracy, $0.429$ test accuracy). Through effectively reducing operator variability and integrating multi-source information, our framework establishes a reliable non-invasive approach for clinical ICP evaluation, holding promise for improving patient management in acute neurological conditions.

cs.CV↗

arXiv2025

Discrepancy-Aware Contrastive Adaptation in Medical Time Series Analysis

In medical time series disease diagnosis, two key challenges are identified. First, the high annotation cost of medical data leads to overfitting in models trained on label-limited, single-center datasets. To address this, we propose incorporating external data from related tasks and leveraging AE-GAN to extract prior knowledge, providing valuable references for downstream tasks. Second, many existing studies employ contrastive learning to derive more generalized medical sequence representations for diagnostic tasks, usually relying on manually designed diverse positive and negative sample pairs. However, these approaches are complex, lack generalizability, and fail to adaptively capture disease-specific features across different conditions. To overcome this, we introduce LMCF (Learnable Multi-views Contrastive Framework), a framework that integrates a multi-head attention mechanism and adaptively learns representations from different views through inter-view and intra-view contrastive learning strategies. Additionally, the pre-trained AE-GAN is used to reconstruct discrepancies in the target data as disease probabilities, which are then integrated into the contrastive learning process. Experiments on three target datasets demonstrate that our method consistently outperforms other seven baselines, highlighting its significant impact on healthcare applications such as the diagnosis of myocardial infarction, Alzheimer's disease, and Parkinson's disease. We release the source code at xxxxx.

cs.HC↗

arXiv2025

Optimizing Vision-Language Consistency via Cross-Layer Regional Attention Alignment

Vision Language Models (VLMs) face challenges in effectively coordinating diverse attention mechanisms for cross-modal embedding learning, leading to mismatched attention and suboptimal performance. We propose Consistent Cross-layer Regional Alignment (CCRA), which introduces Layer-Patch-wise Cross Attention (LPWCA) to capture fine-grained regional-semantic correlations by jointly weighting patch and layer-wise embedding, and Progressive Attention Integration (PAI) that systematically coordinates LPWCA, layer-wise, and patch-wise attention mechanisms in sequence. This progressive design ensures consistency from semantic to regional levels while preventing attention drift and maximizing individual attention benefits. Experimental results on ten diverse vision-language benchmarks demonstrate that our CCRA-enhanced LLaVA-v1.5-7B model achieves state-of-the-art performance, outperforming all baseline methods with only 3.55M additional parameters, while providing enhanced interpretability through more regionally focused and semantically aligned attention patterns.

cs.CV↗

arXiv2025

CARTS: Cooperative and Adaptive Resource Triggering and Stitching for 5G ISAC

This paper presents CARTS, an adaptive 5G uplink sensing scheme designed to provide Integrated Sensing and Communication (ISAC) services. The performance of both communication and sensing fundamentally depends on the availability of accurate and up-to-date channel state information (CSI). In modern 5G networks, uplink CSI is derived from two reference signals: the demodulation reference signal (DMRS) and the sounding reference signal (SRS). However, current base station implementations treat these CSI measurements as separate information streams. The key innovation of CARTS is to fuse these two CSI streams, thereby increasing the frequency of CSI updates and extending sensing opportunities to more users. CARTS addresses two key challenges: (i) a novel channel stitching and compensation method that integrates asynchronous CSI estimates from DMRS and SRS, despite their different time and frequency allocations, and (ii) a real-time SRS triggering algorithm that complements the inherently uncontrollable DMRS schedule, ensuring sufficient and non-redundant sensing opportunities for all users. Our trace-driven evaluation shows that CARTS significantly improves scalability, achieving a channel estimation error (NMSE) of 0.167 and UE tracking accuracy of 85 cm while supporting twice the number of users as a periodic SRS-only baseline with similar performance. By opportunistically combining DMRS and SRS, CARTS therefore provides a practical, standard-compliant solution to improve CSI availability for ISAC without requiring additional radio resources.

cs.NI↗

arXiv2025

A Comprehensive Survey for Real-World Industrial Defect Detection: Challenges, Approaches, and Prospects

Industrial defect detection is vital for upholding product quality across contemporary manufacturing systems. As the expectations for precision, automation, and scalability intensify, conventional inspection approaches are increasingly found wanting in addressing real-world demands. Notable progress in computer vision and deep learning has substantially bolstered defect detection capabilities across both 2D and 3D modalities. A significant development has been the pivot from closed-set to open-set defect detection frameworks, which diminishes the necessity for extensive defect annotations and facilitates the recognition of novel anomalies. Despite such strides, a cohesive and contemporary understanding of industrial defect detection remains elusive. Consequently, this survey delivers an in-depth analysis of both closed-set and open-set defect detection strategies within 2D and 3D modalities, charting their evolution in recent years and underscoring the rising prominence of open-set techniques. We distill critical challenges inherent in practical detection environments and illuminate emerging trends, thereby providing a current and comprehensive vista of this swiftly progressing field.

cs.CV↗

arXiv2025

Intelligent Histology for Tumor Neurosurgery

The importance of rapid and accurate histologic analysis of surgical tissue in the operating room has been recognized for over a century. Our standard-of-care intraoperative pathology workflow is based on light microscopy and H\&E histology, which is slow, resource-intensive, and lacks real-time digital imaging capabilities. Here, we present an emerging and innovative method for intraoperative histologic analysis, called Intelligent Histology, that integrates artificial intelligence (AI) with stimulated Raman histology (SRH). SRH is a rapid, label-free, digital imaging method for real-time microscopic tumor tissue analysis. SRH generates high-resolution digital images of surgical specimens within seconds, enabling AI-driven tumor histologic analysis, molecular classification, and tumor infiltration detection. We review the scientific background, clinical translation, and future applications of intelligent histology in tumor neurosurgery. We focus on the major scientific and clinical studies that have demonstrated the transformative potential of intelligent histology across multiple neurosurgical specialties, including neurosurgical oncology, skull base, spine oncology, pediatric tumors, and periperal nerve tumors. Future directions include the development of AI foundation models through multi-institutional datasets, incorporating clinical and radiologic data for multimodal learning, and predicting patient outcomes. Intelligent histology represents a transformative intraoperative workflow that can reinvent real-time tumor analysis for 21st century neurosurgery.

cs.CV↗