arXiv subjects

Yuhan Zhou

Publications and source records attributed to Yuhan Zhou.

17 recordsLinked to original sources

arXiv2026

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

Recursive self-improvement (RSI) enables AI systems to turn experience and feedback into persistent changes that improve both their capabilities and the process of future improvement. We first use the Headroom-Closed Index (HCI) to reveal the problems of existing LLMs, then introduce the RSI concept and its development roadmap: from improvement-execution autonomy, improvement-strategy autonomy, experience-acquisition autonomy, and environment-adaptation autonomy, to recursive meta-improvement. Next we examine RSI across scenarios (e.g., scientific discovery, embodied intelligence, software engineering), highlighting their distinct requirements and development speeds. Drawing on diverse industry practices and preliminary empirical evidence, we connect RSI research with practical systems and identify key challenges to achieving genuine RSI.

cs.LG↗

arXiv2026

Dissecting Neuro-Symbolic Quality Assurance for Synthetic Oncology Data Generation

Synthetic clinical data generation with large language models addresses the scarcity that limits cancer staging research, but oncology hallucinations are categorically harmful: one clinically impossible staging assignment contaminates every downstream model trained on it. Neuro-symbolic pipelines validate during generation, yet the contribution of individual quality-assurance components remains unclear. We report three controlled studies isolating gate necessity, constraint attribution, and retrieval conditionality, holding generation protocol, diversity thresholds, and fine-tuning hyperparameters constant across adapter conditions. The symbolic gate enforces schema completeness, ontology coverage against the Systematized Nomenclature of Medicine, and staging-logic consistency under American Joint Committee on Cancer eighth-edition rules. Ungated, 29.9% of records contain schema failures and 20.1% contain clinically invalid staging. Schema validation is the load-bearing filter: within the fully gated corpus it rejects 148 of 512 records, ontology grounding a further 24, and staging-logic validation none---the only generator producing logic violations is already excluded on schema, making clinical-logic validation a generator-conditional safeguard rather than the dominant filter. Retrieval augmentation is strongly model-dependent: it improves gate compliance for one generator by 12.5 percentage points, has no measurable effect for a second, and collapses output in a third. Across gated configurations ontology density is largely unchanged, indicating that symbolic validation improves clinical validity rather than vocabulary richness. Symbolic gating therefore buys corpus validity but no commensurate gain on real lung-cancer notes in this study; retrieval should be evaluated per model, and ontology density should not be reported as a proxy for corpus quality.

cs.AI↗

arXiv2026

TensorCast: The Missing Tensor Management Layer in Large Language Model Infrastructure

Modern LLM infrastructure increasingly manages tensors not only as computation data, but also as persistent states shared across distributed components. Existing systems optimize individual tensor management tasks, such as model weight loading, KV cache management, and checkpoint synchronization, by deeply integrating task-specific mechanisms with execution engines, networks, or storage backends. However, this specialization creates isolated silos that hinder the reuse and composition of tensor management strategies across evolving LLM workloads. In this paper, we identify tensor lifecycle management as a missing abstraction layer in LLM infrastructure and propose Tensor-as-a-Service (TaaS), which decouples tensor state management from computation logic. We design and build TensorCast, a distributed tensor management layer that provides first-class tensor abstractions, programmable lifecycle primitives, and a runtime that separates tensor management policies from execution mechanisms. This enables developers to write tensor management programs using TensorCast APIs while transparently leveraging distributed execution and data movement. We integrate TensorCast with vLLM and SGLang and evaluate it across diverse tensor lifecycle workloads, including model weight materialization, weight synchronization, KV cache management, and programmable request routing. Our results show that TensorCast achieves competitive performance with specialized tensor management systems while enabling new cross-component optimization policies. A programmable policy implemented with TensorCast improves median TTFT by up to 93.2% under highly concurrent multi-turn agent workloads.

cs.DC↗

arXiv2026

Neuromorphic Speech Enhancement with Dual-Branch Spiking Neural Networks

Spiking neural network (SNN)-based neuromorphic speech enhancement has emerged as a promising paradigm due to its energy efficiency, yet it still underperforms classical artificial neural network (ANN)-based approaches owing to binary activations and the lack of well-designed network architectures. To overcome this limitation, we propose a novel dual-branch spiking neural network architecture equipped with a gated spiking unit (GSU), termed GSU-DBNet. Specifically, GSU-DBNet simultaneously models the speech magnitude spectrum and complex spectrum, predicting the corresponding magnitude and complex spectral masks. Meanwhile, a dual-path GSU module is adopted to exploit temporal and frequency information for enhanced spatiotemporal feature representation. Experiments on a popular benchmark dataset show that GSU-DBNet achieves a PESQ score of 3.04 with only 394K parameters, outperforming existing SNN-based methods while using only 4.5%--10.6% of the parameters of representative ANN-based models.

cs.SD↗

arXiv2026

PassNet: Scaling Large Language Models for Graph Compiler Pass Generation

Modern tensor compilers such as TorchInductor deliver substantial speedups on mainstream models, yet face a systematic performance ceiling on long-tail workloads -- our profiling shows that 43% of real-world subgraphs experience end-to-end slowdowns under default compilation. While LLMs offer a path toward automated optimization, existing efforts focus on standalone kernel generation. We argue that pass generation -- where LLMs author structured graph transformations that integrate directly into compiler pipelines -- is the more appropriate abstraction. We propose PassNet, the first large-scale ecosystem for LLM-based compiler pass generation, comprising: (1) PassNet-Dataset, over 18K unique computational graphs from 100K real-world models; and (2) PassBench, 200 curated long-tail fusible tasks (comprising 2,060 subgraphs in total) evaluated under the Error-aware Speedup Score (ES_t) -- a metric unifying correctness, stability, and performance -- with layered integrity defenses against systematic LLM exploitation. Experiments reveal that PassBench is both highly discriminative and genuinely unsaturated: the best frontier model trails TorchInductor by 37% in aggregate, yet on individual subgraphs LLMs achieve up to 3x speedup over the same compiler -- indicating that the bottleneck is consistency, not capability. Fine-tuning a small model on merely ~4K PassNet trajectories yields a 2.67x improvement approaching frontier-model performance, demonstrating substantial headroom and validating PassNet as live training infrastructure for advancing LLM-driven compiler optimization. All data, benchmarks, and tooling are publicly available.

cs.AI↗

arXiv2026

BLADE: Adaptive Wi-Fi Contention Control for Next-Generation Real-Time Communication

Next-generation real-time communication (NGRTC) applications, such as cloud gaming and XR, demand consistently ultra-low latency. However, through our first large-scale measurement, we find that despite the deployment of edge servers, dedicated congestion control, and loss recovery mechanisms, cloud gaming users still experience long-tail latency in Wi-Fi networks. We further identify that Wi-Fi last-mile access points (APs) serve as the primary latency bottleneck. Specifically, short-term packet delivery droughts, caused by fundamental limitations in Wi-Fi contention control standards, are the root cause. To address this issue, we propose BLADE, an adaptive contention control algorithm that dynamically adjusts the contention windows (CW) of all Wi-Fi transmitters based on the channel contention level in a fully distributed manner. Our NS3 simulations and real-world evaluations with commercial Wi-Fi APs demonstrate that, compared to standard contention control, BLADE reduces Wi-Fi packet transmission tail latency by over 5X under heavy channel contention and significantly stabilizes MAC throughput while ensuring fast and fair convergence. Consequently, BLADE reduces the video stall rate in cloud gaming by over 90%.

cs.NI↗

arXiv2026

Modeling and Measuring Redundancy in Multisource Multimodal Data for Autonomous Driving

Next-generation autonomous vehicles (AVs) rely on large volumes of multisource and multimodal ($M^2$) data to support real-time decision-making. In practice, data quality (DQ) varies across sources and modalities due to environmental conditions and sensor limitations, yet AV research has largely prioritized algorithm design over DQ analysis. This work focuses on redundancy as a fundamental but underexplored DQ issue in AV datasets. Using the nuScenes and Argoverse 2 (AV2) datasets, we model and measure redundancy in multisource camera data and multimodal image-LiDAR data, and evaluate how removing redundant labels affects the YOLOv8 object detection task. Experimental results show that selectively removing redundant multisource image object labels from cameras with shared fields of view improves detection. In nuScenes, mAP${50}$ gains from $0.66$ to $0.70$, $0.64$ to $0.67$, and from $0.53$ to $0.55$, on three representative overlap regions, while detection on other overlapping camera pairs remains at the baseline even under stronger pruning. In AV2, $4.1$-$8.6\%$ of labels are removed, and mAP${50}$ stays near the $0.64$ baseline. Multimodal analysis also reveals substantial redundancy between image and LiDAR data. These findings demonstrate that redundancy is a measurable and actionable DQ factor with direct implications for AV performance. This work highlights the role of redundancy as a data quality factor in AV perception and motivates a data-centric perspective for evaluating and improving AV datasets. Code, data, and implementation details are publicly available at: https://github.com/yhZHOU515/RedundancyAD

cs.CV↗

arXiv2026

Is He Extroverted? Identifying Missing Relevant Personas for Faithful User Simulation

Existing user simulation approaches focus on generating user-like responses in dialogue. They often assume that the provided persona is sufficient for producing such responses, without verifying whether critical personas are supplied. This raises concerns about the validity of simulation results. To address this issue, we study the task of identifying persona dimensions (e.g., "whether the user is price-sensitive") that are relevant but missing in simulating a user's reply for a given dialogue context. We introduce PICQ-drama (constructed from TVShowGuess), a benchmark of context-aware choice questions, annotated with missing persona dimensions whose absence leads to ambiguous user choices. We further design diverse evaluation criteria for missing persona identification. Benchmarking leading LLMs on our PICQ-drama dataset demonstrates the feasibility of this task. Evaluation across diverse criteria, along with further analyses, reveals cognitive differences between LLMs and humans and highlights the distinct roles of different persona categories in shaping responses.

cs.HC↗

arXiv2025

Efficient Rectified Flow for Image Fusion

Image fusion is a fundamental and important task in computer vision, aiming to combine complementary information from different modalities to fuse images. In recent years, diffusion models have made significant developments in the field of image fusion. However, diffusion models often require complex computations and redundant inference time, which reduces the applicability of these methods. To address this issue, we propose RFfusion, an efficient one-step diffusion model for image fusion based on Rectified Flow. We incorporate Rectified Flow into the image fusion task to straighten the sampling path in the diffusion model, achieving one-step sampling without the need for additional training, while still maintaining high-quality fusion results. Furthermore, we propose a task-specific variational autoencoder (VAE) architecture tailored for image fusion, where the fusion operation is embedded within the latent space to further reduce computational complexity. To address the inherent discrepancy between conventional reconstruction-oriented VAE objectives and the requirements of image fusion, we introduce a two-stage training strategy. This approach facilitates the effective learning and integration of complementary information from multi-modal source images, thereby enabling the model to retain fine-grained structural details while significantly enhancing inference efficiency. Extensive experiments demonstrate that our method outperforms other state-of-the-art methods in terms of both inference speed and fusion quality. Code is available at https://github.com/zirui0625/RFfusion.

cs.CV↗

arXiv2025

Disaster Informatics after the COVID-19 Pandemic: Bibliometric and Topic Analysis based on Large-scale Academic Literature

This study presents a comprehensive bibliometric and topic analysis of the disaster informatics literature published between January 2020 to September 2022. Leveraging a large-scale corpus and advanced techniques such as pre-trained language models and generative AI, we identify the most active countries, institutions, authors, collaboration networks, emergent topics, patterns among the most significant topics, and shifts in research priorities spurred by the COVID-19 pandemic. Our findings highlight (1) countries that were most impacted by the COVID-19 pandemic were also among the most active, with each country having specific research interests, (2) countries and institutions within the same region or share a common language tend to collaborate, (3) top active authors tend to form close partnerships with one or two key partners, (4) authors typically specialized in one or two specific topics, while institutions had more diverse interests across several topics, and (5) the COVID-19 pandemic has influenced research priorities in disaster informatics, placing greater emphasis on public health. We further demonstrate that the field is converging on multidimensional resilience strategies and cross-sectoral data-sharing collaborations or projects, reflecting a heightened awareness of global vulnerability and interdependency. Collecting and quality assurance strategies, data analytic practices, LLM-based topic extraction and summarization approaches, and result visualization tools can be applied to comparable datasets or solve similar analytic problems. By mapping out the trends in disaster informatics, our analysis offers strategic insights for policymakers, practitioners, and scholars aiming to enhance disaster informatics capacities in an increasingly uncertain and complex risk landscape.

cs.SI↗

arXiv2025

A Novel Multi-layer Task-centric and Data Quality Framework for Autonomous Driving

The next-generation autonomous vehicles (AVs), embedded with frequent real-time decision-making, will rely heavily on a large volume of multisource and multimodal data. In real-world settings, the data quality (DQ) of different sources and modalities usually varies due to unexpected environmental factors or sensor issues. However, both researchers and practitioners in the AV field overwhelmingly concentrate on models/algorithms while undervaluing the DQ. To fulfill the needs of the next-generation AVs with guarantees of functionality, efficiency, and trustworthiness, this paper proposes a novel task-centric and data quality vase framework which consists of five layers: data layer, DQ layer, task layer, application layer, and goal layer. The proposed framework aims to map DQ with task requirements and performance goals. To illustrate, a case study investigating redundancy on the nuScenes dataset proves that partially removing redundancy on multisource image data could improve YOLOv8 object detection task performance. Analysis on multimodal data of image and LiDAR further presents existing redundancy DQ issues. This paper opens up a range of critical but unexplored challenges at the intersection of DQ, task orchestration, and performance-oriented system development in AVs. It is expected to guide the AV community toward building more adaptive, explainable, and resilient AVs that respond intelligently to dynamic environments and heterogeneous data streams. Code, data, and implementation details are publicly available at: https://anonymous.4open.science/r/dq4av-framework/README.md.

cs.CV↗

arXiv2025

Periapsis shift in magnetized stationary and axisymmetric spacetimes

In this work, we conduct a detailed study of the precession of charged particles in stationary and asymmetric spacetimes with external magnetic fields. Specifically, we develop the post-Newtonian method and the quasi-circular approximation to derive the periapsis shift respectively for two common types of magnetic fields, the dipolar one and the asymptotically uniform one. It is found using the PN method for magnetic fields decaying as fast or faster than a magnetic dipole that the magnetic effect in the periapsis shift appears from the same order of the traditional frame-dragging term due to the spacetime spin. The magnetic field is found to enhance (or decrease) the periapsis shift when the Lorentz force is attractive (or repulsive). When the repulsive Lorentz force is strong enough, the periapsis shift can become negative. For magnetic fields with a slower decay rate, the periapsis shift of quasi-circular orbits exhibits a more complex dependence on the Lorentz force. The periapsis shift increases as the attractive Lorentz force increases from zero but will decrease eventually. However, when the Lorentz force is repulsive, the orbit develops local helical loops and the periapsis shift approaches $-2π$. The results for the dipolar magnetic field are applied to the periapsis shift of Mercury around the Sun and the S2 around the Sgr A* to constrain the dipole moment of the center and the charge of the orbiting object.

gr-qc↗

arXiv2024

A Survey on Data Quality Dimensions and Tools for Machine Learning

Machine learning (ML) technologies have become substantial in practically all aspects of our society, and data quality (DQ) is critical for the performance, fairness, robustness, safety, and scalability of ML models. With the large and complex data in data-centric AI, traditional methods like exploratory data analysis (EDA) and cross-validation (CV) face challenges, highlighting the importance of mastering DQ tools. In this survey, we review 17 DQ evaluation and improvement tools in the last 5 years. By introducing the DQ dimensions, metrics, and main functions embedded in these tools, we compare their strengths and limitations and propose a roadmap for developing open-source DQ tools for ML. Based on the discussions on the challenges and emerging trends, we further highlight the potential applications of large language models (LLMs) and generative AI in DQ evaluation and improvement for ML. We believe this comprehensive survey can enhance understanding of DQ in ML and could drive progress in data-centric AI. A complete list of the literature investigated in this survey is available on GitHub at: https://github.com/haihua0913/awesome-dq4ml.

cs.LG↗

arXiv2021

Multi-View Spatial-Temporal Graph Convolutional Networks with Domain Generalization for Sleep Stage Classification

Sleep stage classification is essential for sleep assessment and disease diagnosis. Although previous attempts to classify sleep stages have achieved high classification performance, several challenges remain open: 1) How to effectively utilize time-varying spatial and temporal features from multi-channel brain signals remains challenging. Prior works have not been able to fully utilize the spatial topological information among brain regions. 2) Due to the many differences found in individual biological signals, how to overcome the differences of subjects and improve the generalization of deep neural networks is important. 3) Most deep learning methods ignore the interpretability of the model to the brain. To address the above challenges, we propose a multi-view spatial-temporal graph convolutional networks (MSTGCN) with domain generalization for sleep stage classification. Specifically, we construct two brain view graphs for MSTGCN based on the functional connectivity and physical distance proximity of the brain regions. The MSTGCN consists of graph convolutions for extracting spatial features and temporal convolutions for capturing the transition rules among sleep stages. In addition, attention mechanism is employed for capturing the most relevant spatial-temporal information for sleep stage classification. Finally, domain generalization and MSTGCN are integrated into a unified framework to extract subject-invariant sleep features. Experiments on two public datasets demonstrate that the proposed model outperforms the state-of-the-art baselines.

eess.SP↗

arXiv2017

Fronthaul Compression and Transmit Beamforming Optimization for Multi-Antenna Uplink C-RAN

This paper considers the joint fronthaul compression and transmit beamforming design for the uplink cloud radio access network (C-RAN), in which multi-antenna user terminals communicate with a cloud-computing based centralized processor (CP) through multi-antenna base-stations (BSs) serving as relay nodes. A compress-and-forward relaying strategy, named the VMAC scheme, is employed, in which the BSs can either perform single-user compression or Wyner-Ziv coding to quantize the received signals and send the quantization bits to the CP via capacity-limited fronthaul links; the CP performs successive decoding with either successive interference cancellation (SIC) receiver or linear minimum-mean-square-error (MMSE) receiver. Under this setup, this paper investigates the joint optimization of the transmit beamformers at the users and the quantization noise covariance matrices at the BSs for maximizing the network utility. A novel weighted minimum-mean-square-error successive convex approximation (WMMSE-SCA) algorithm is first proposed for maximizing the weighted sum rate under the user transmit power and fronthaul capacity constraints with single-user compression. Assuming a heuristic decompression order, the proposed algorithm is then adapted for optimizing the transmit beamforming and fronthaul compression under Wyner-Ziv coding. This paper also proposes a low-complexity separate design consisting of optimizing transmit beamformers for the Gaussian vector multiple-access channel along with per-antenna quantizers with uniform quantization noise levels across the antennas at each BS. Numerical results show that with optimized beamforming and fronthaul compression, C-RAN can significantly outperform conventional cellular networks. Furthermore, the low complexity separate design already performs very close to the optimized joint design in regime of practical interest.

cs.IT↗

arXiv2016

On Optimal Fronthaul Compression and Decoding Strategies for Uplink Cloud Radio Access Networks

This paper investigates the compress-and-forward scheme for an uplink cloud radio access network (C-RAN) model, where multi-antenna base-stations (BSs) are connected to a cloud-computing based central processor (CP) via capacity-limited fronthaul links. The BSs compress the received signals with Wyner-Ziv coding and send the representation bits to the CP; the CP performs the decoding of all the users' messages. Under this setup, this paper makes progress toward the optimal structure of the fronthaul compression and CP decoding strategies for the compress-and-forward scheme in C-RAN. On the CP decoding strategy design, this paper shows that under a sum fronthaul capacity constraint, a generalized successive decoding strategy of the quantization and user message codewords that allows arbitrary interleaved order at the CP achieves the same rate region as the optimal joint decoding. Further, it is shown that a practical strategy of successively decoding the quantization codewords first, then the user messages, achieves the same maximum sum rate as joint decoding under individual fronthaul constraints. On the joint optimization of user transmission and BS quantization strategies, this paper shows that if the input distributions are assumed to be Gaussian, then under joint decoding, the optimal quantization scheme for maximizing the achievable rate region is Gaussian. Moreover, Gaussian input and Gaussian quantization with joint decoding achieve to within a constant gap of the capacity region of the Gaussian multiple-input multiple-output (MIMO) uplink C-RAN model. Finally, this paper addresses the computational aspect of optimizing uplink MIMO C-RAN by showing that under fixed Gaussian input, the sum rate maximization problem over the Gaussian quantization noise covariance matrices can be formulated as convex optimization problems, thereby facilitating its efficient solution.

cs.IT↗

arXiv2014

Optimized Backhaul Compression for Uplink Cloud Radio Access Network

This paper studies the uplink of a cloud radio access network (C-RAN) where the cell sites are connected to a cloud-computing-based central processor (CP) with noiseless backhaul links with finite capacities. We employ a simple compress-and-forward scheme in which the base-stations(BSs) quantize the received signals and send the quantized signals to the CP using either distributed Wyner-Ziv coding or single-user compression. The CP decodes the quantization codewords first, then decodes the user messages as if the remote users and the cloud center form a virtual multiple-access channel (VMAC). This paper formulates the problem of optimizing the quantization noise levels for weighted sum rate maximization under a sum backhaul capacity constraint. We propose an alternating convex optimization approach to find a local optimum solution to the problem efficiently, and more importantly, establish that setting the quantization noise levels to be proportional to the background noise levels is near optimal for sum-rate maximization when the signal-to-quantization-noise ratio (SQNR) is high. In addition, with Wyner-Ziv coding, the approximate quantization noise level is shown to achieve the sum-capacity of the uplink C-RAN model to within a constant gap. With single-user compression, a similar constant-gap result is obtained under a diagonal dominant channel condition. These results lead to an efficient algorithm for allocating the backhaul capacities in C-RAN. The performance of the proposed scheme is evaluated for practical multicell and heterogeneous networks. It is shown that multicell processing with optimized quantization noise levels across the BSs can significantly improve the performance of wireless cellular networks.

cs.IT↗