Searcharxiv⌕ Search

arXiv subjects

Yu Yan

Publications and source records attributed to Yu Yan.

At least 55 records · Page 3Linked to original sources
arXiv2025

Sigma: Differential Rescaling of Query, Key and Value for Efficient Language Models

We introduce Sigma, an efficient large language model specialized for the system domain, empowered by a novel architecture including DiffQKV attention, and pre-trained on our meticulously collected system domain data. DiffQKV attention significantly enhances the inference efficiency of Sigma by optimizing the Query (Q), Key (K), and Value (V) components in the attention mechanism differentially, based on their varying impacts on the model performance and efficiency indicators. Specifically, we (1) conduct extensive experiments that demonstrate the model's varying sensitivity to the compression of K and V components, leading to the development of differentially compressed KV, and (2) propose augmented Q to expand the Q head dimension, which enhances the model's representation capacity with minimal impacts on the inference speed. Rigorous theoretical and empirical analyses reveal that DiffQKV attention significantly enhances efficiency, achieving up to a 33.36% improvement in inference speed over the conventional grouped-query attention (GQA) in long-context scenarios. We pre-train Sigma on 6T tokens from various sources, including 19.5B system domain data that we carefully collect and 1T tokens of synthesized and rewritten data. In general domains, Sigma achieves comparable performance to other state-of-arts models. In the system domain, we introduce the first comprehensive benchmark AIMicius, where Sigma demonstrates remarkable performance across all tasks, significantly outperforming GPT-4 with an absolute improvement up to 52.5%.

cs.CL↗
arXiv2024

Xmodel-LM Technical Report

We introduce Xmodel-LM, a compact and efficient 1.1B language model pre-trained on around 2 trillion tokens. Trained on our self-built dataset (Xdata), which balances Chinese and English corpora based on downstream task optimization, Xmodel-LM exhibits remarkable performance despite its smaller size. It notably surpasses existing open-source language models of similar scale. Our model checkpoints and code are publicly accessible on GitHub at https://github.com/XiaoduoAILab/XmodelLM.

cs.CL↗
arXiv2024

InstruGen: Automatic Instruction Generation for Vision-and-Language Navigation Via Large Multimodal Models

Recent research on Vision-and-Language Navigation (VLN) indicates that agents suffer from poor generalization in unseen environments due to the lack of realistic training environments and high-quality path-instruction pairs. Most existing methods for constructing realistic navigation scenes have high costs, and the extension of instructions mainly relies on predefined templates or rules, lacking adaptability. To alleviate the issue, we propose InstruGen, a VLN path-instruction pairs generation paradigm. Specifically, we use YouTube house tour videos as realistic navigation scenes and leverage the powerful visual understanding and generation abilities of large multimodal models (LMMs) to automatically generate diverse and high-quality VLN path-instruction pairs. Our method generates navigation instructions with different granularities and achieves fine-grained alignment between instructions and visual observations, which was difficult to achieve with previous methods. Additionally, we design a multi-stage verification mechanism to reduce hallucinations and inconsistency of LMMs. Experimental results demonstrate that agents trained with path-instruction pairs generated by InstruGen achieves state-of-the-art performance on the R2R and RxR benchmarks, particularly in unseen environments. Code is available at https://github.com/yanyu0526/InstruGen.

cs.RO↗
arXiv2024

CliBench: A Multifaceted and Multigranular Evaluation of Large Language Models for Clinical Decision Making

The integration of Artificial Intelligence (AI), especially Large Language Models (LLMs), into the clinical diagnosis process offers significant potential to improve the efficiency and accessibility of medical care. While LLMs have shown some promise in the medical domain, their application in clinical diagnosis remains underexplored, especially in real-world clinical practice, where highly sophisticated, patient-specific decisions need to be made. Current evaluations of LLMs in this field are often narrow in scope, focusing on specific diseases or specialties and employing simplified diagnostic tasks. To bridge this gap, we introduce CliBench, a novel benchmark developed from the MIMIC IV dataset, offering a comprehensive and realistic assessment of LLMs' capabilities in clinical diagnosis. This benchmark not only covers diagnoses from a diverse range of medical cases across various specialties but also incorporates tasks of clinical significance: treatment procedure identification, lab test ordering and medication prescriptions. Supported by structured output ontologies, CliBench enables a precise and multi-granular evaluation, offering an in-depth understanding of LLM's capability on diverse clinical tasks of desired granularity. We conduct a zero-shot evaluation of leading LLMs to assess their proficiency in clinical decision-making. Our preliminary results shed light on the potential and limitations of current LLMs in clinical settings, providing valuable insights for future advancements in LLM-powered healthcare.

cs.CL↗
arXiv2024

Building an Ethical and Trustworthy Biomedical AI Ecosystem for the Translational and Clinical Integration of Foundational Models

Foundational Models (FMs) are gaining increasing attention in the biomedical AI ecosystem due to their ability to represent and contextualize multimodal biomedical data. These capabilities make FMs a valuable tool for a variety of tasks, including biomedical reasoning, hypothesis generation, and interpreting complex imaging data. In this review paper, we address the unique challenges associated with establishing an ethical and trustworthy biomedical AI ecosystem, with a particular focus on the development of FMs and their downstream applications. We explore strategies that can be implemented throughout the biomedical AI pipeline to effectively tackle these challenges, ensuring that these FMs are translated responsibly into clinical and translational settings. Additionally, we emphasize the importance of key stewardship and co-design principles that not only ensure robust regulation but also guarantee that the interests of all stakeholders, especially those involved in or affected by these clinical and translational applications are adequately represented. We aim to empower the biomedical AI community to harness these models responsibly and effectively. As we navigate this exciting frontier, our collective commitment to ethical stewardship, co-design, and responsible translation will be instrumental in ensuring that the evolution of FMs truly enhances patient care and medical decision making, ultimately leading to a more equitable and trustworthy biomedical AI ecosystem.

cs.CY↗
arXiv2024

Explainable Biomedical Hypothesis Generation via Retrieval Augmented Generation enabled Large Language Models

The vast amount of biomedical information available today presents a significant challenge for investigators seeking to digest, process, and understand these findings effectively. Large Language Models (LLMs) have emerged as powerful tools to navigate this complex and challenging data landscape. However, LLMs may lead to hallucinatory responses, making Retrieval Augmented Generation (RAG) crucial for achieving accurate information. In this protocol, we present RUGGED (Retrieval Under Graph-Guided Explainable disease Distinction), a comprehensive workflow designed to support investigators with knowledge integration and hypothesis generation, identifying validated paths forward. Relevant biomedical information from publications and knowledge bases are reviewed, integrated, and extracted via text-mining association analysis and explainable graph prediction models on disease nodes, forecasting potential links among drugs and diseases. These analyses, along with biomedical texts, are integrated into a framework that facilitates user-directed mechanism elucidation as well as hypothesis exploration through RAG-enabled LLMs. A clinical use-case demonstrates RUGGED's ability to evaluate and recommend therapeutics for Arrhythmogenic Cardiomyopathy (ACM) and Dilated Cardiomyopathy (DCM), analyzing prescribed drugs for molecular interactions and unexplored uses. The platform minimizes LLM hallucinations, offers actionable insights, and improves the investigation of novel therapeutics.

cs.CL↗
arXiv2024

Out-of-Plane Polarization from Spin Reflection Induces Field-Free Spin-Orbit Torque Switching in Structures with Canted NiO Interfacial Moments

Realizing deterministic current-induced spin-orbit torque (SOT) magnetization switching, especially in systems exhibiting perpendicular magnetic anisotropy (PMA), typically requires the application of a collinear in-plane field, posing a challenging problem. In this study, we successfully achieve field-free SOT switching in the CoFeB/MgO system. In a Ta/CoFeB/MgO/NiO/Ta structure, spin reflection at the NiO interface, characterized by noncollinear spin structures with canted magnetization, generates a spin current with an out-of-plane spin polarization σz. We confirm the contribution of σz to the field-free SOT switching through measurements of the shift effect in the out-of-plane magnetization hysteresis loops under different currents. The incorporation of NiO as an antiferromagnetic insulator, mitigates the current shunting effect and ensures excellent thermal stability of the device. The sample with 0.8 nm MgO and 2 nm NiO demonstrates an impressive optimal switching ratio approaching 100% without an in-plane field. This breakthrough in the CoFeB/MgO system promises significant applications in spintronics, advancing us closer to realizing innovative technologies.

physics.app-ph↗
arXiv2024

KnobCF: Uncertainty-aware Knob Tuning

The knob tuning aims to optimize database performance by searching for the most effective knob configuration under a certain workload. Existing works suffer two significant problems. On the one hand, there exist multiple similar even useless evaluations of knob tuning even with the diverse searching methods because of the different sensitivities of knobs on a certain workload. On the other hand, the single evaluation of knob configurations may bring overestimation or underestimation because of the query uncertainty performance. To solve the above problems, we propose a decoupled query uncertainty-aware knob classifier, called KnobCF, to enhance the knob tuning. Our method has three significant contributions: (1) We propose a novel concept of the uncertainty-aware knob configuration estimation to enhance the knob tuning process. (2) We provide an effective few-shot uncertainty knob estimator without extra time consumption in training data collection, which has a high time efficiency in practical tuning tasks. (3) Our method provides a general framework that could be easily deployed in any knob tuning task because we make no changes to the knob tuners and the database management system. Our experiments on four open-source benchmarks demonstrate that our method effectively reduces useless evaluations and improves the tuning results. Especially in TPCC, our method achieves competitive tuning results with only 60% to 70% time consumption compared to the full workload evaluations.

cs.DB↗
arXiv2024

EMIT: Micro-Invasive Database Configuration Tuning

The process of database knob tuning has always been a challenging task. Recently, database knob tuning methods has emerged as a promising solution to mitigate these issues. However, these methods still face certain limitations.On one hand, when applying knob tuning algorithms to optimize databases in practice, it either requires frequent updates to the database or necessitates acquiring database workload and optimizing through workload replay. The former approach involves constant exploration and updating of database configurations, inevitably leading to a decline in database performance during optimization. The latter, on the other hand, requires the acquisition of workload data, which could lead to data leakage issues. Moreover, the hyperparameter configuration space for database knobs is vast, making it challenging for optimizers to converge. These factors significantly hinder the practical implementation of database tuning. To address these concerns, we proposes an efficient and micro-invasive knob tuning method. This method relies on workload synthesis on cloned databases to simulate the workload that needs tuning, thus minimizing the intrusion on the database. And we utilizing a configuration replacement strategy to filter configuration candidates that perform well under the synthesized workload to find best configuration. And during the tuning process, we employ a knowledge transfer method to extract a common high-performance space, to boost the convergence of the optimizer.

cs.DB↗
arXiv2024

Harmonic and Interharmonic Detection in Power Systems Based on Fractal-Optimized Variational Mode Decomposition

The proposed method introduces a parameter determination approach based on the minimum Fractal box dimension (FBD) of Variational Mode Decomposition (VMD) components, aiming to address the issue of manual determination of VMD decomposition layers in advance. Initially, VMD is applied to the original power signal, and the layer number for VMD decomposition is determined by selecting the K value associated with the smallest fractal box dimension among its components. Subsequently, several Intrinsic Mode Functions (IMFs) are obtained as fundamental, harmonic, and interharmonic signals representing different aspects of the power system. Furthermore, Hilbert transform(HT) is employed to extract instantaneous amplitude and frequency information from these harmonic signals. Experimental evaluation using simulation data and real-world power system data demonstrates that compared to Empirical Mode Decomposition (EMD) and Ensemble Empirical Mode Decomposition (EEMD), our proposed method achieves more accurate identification and effective extraction of harmonic signals.

eess.SP↗
arXiv2024

Gas pressure manipulation of exciton states in monolayer WS2

Over the past few decades, thin film optoelectronic devices based on transition metal dichalcogenides (TMDs) have made significant progress. However, the sensitivity of the exciton states to environmental change presents challenges for device applications. This work reports on the evolution of photo-induced exciton states in monolayer WS2 in a chamber with low gas pressure. It elucidates the physical mechanism of the transition between neutral and charged excitons. At 222 mTorr, the transition rate between excitons includes two components, 0.09 s-1 and 1.68 s-1, respectively. Based on this phenomenon, we have developed a pressure-tuning method for exciton manipulation, allowing a tuning range of approximately 40% in exciton weight. We also demonstrate that the intensity of neutral exciton emission from monolayer WS2 follows a power-law distribution concerning gas pressure, indicating a highly sensitive pressure dependence. This work presents a non-destructive and highly sensitive method for exciton conversion through in-situ manipulation. It highlights the potential development of monolayer WS2 in pressure sensing and explains the impact of environmental factors on product quality in photovoltaic devices.

physics.optics↗
arXiv2023

Duet: efficient and scalable hybriD neUral rElation undersTanding

Learned cardinality estimation methods have achieved high precision compared to traditional methods. Among learned methods, query-driven approaches have faced the workload drift problem for a long time. Although both data-driven and hybrid methods are proposed to avoid this problem, most of them suffer from high training and estimation costs, limited scalability, instability, and long-tail distribution problems on high-dimensional tables, which seriously affects the practical application of learned cardinality estimators. In this paper, we prove that most of these problems are directly caused by the widely used progressive sampling. We solve this problem by introducing predicate information into the autoregressive model and propose Duet, a stable, efficient, and scalable hybrid method to estimate cardinality directly without sampling or any non-differentiable process, which can not only reduce the inference complexity from $O(n)$ to $O(1)$ compared to Naru and UAE but also achieve higher accuracy on high cardinality and high-dimensional tables. Experimental results show that Duet can achieve all the design goals above and be much more practical. Besides, Duet even has a lower inference cost on CPU than that of most learned methods on GPU.

cs.DB↗
arXiv2023

QCFE: An efficient Feature engineering for query cost estimation

Query cost estimation is a classical task for database management. Recently, researchers apply the AI-driven model to implement query cost estimation for achieving high accuracy. However, two defects of feature design lead to poor cost estimation accuracy-time efficiency. On the one hand, existing works only encode the query plan and data statistics while ignoring some other important variables, like storage structure, hardware, database knobs, etc. These variables also have significant impact on the query cost. On the other hand, due to the straightforward encoding design, existing works suffer heavy representation learning burden on ineffective dimensions of input. To meet the above two problems, we first propose an efficient feature engineering for query cost estimation, called QCFE. Specifically, we design a novel feature called feature snapshot to efficiently integrate the influences of the ignored variables. Further, we propose a difference-propagation feature reduction method for query cost estimation to filter the useless features. The experimental results demonstrate our QCFE could largely improve the time-accuracy efficiency on extensive benchmarks.

cs.DB↗
arXiv2023

Enhancing Mobile Face Anti-Spoofing: A Robust Framework for Diverse Attack Types under Screen Flash

Face anti-spoofing (FAS) is crucial for securing face recognition systems. However, existing FAS methods with handcrafted binary or pixel-wise labels have limitations due to diverse presentation attacks (PAs). In this paper, we propose an attack type robust face anti-spoofing framework under light flash, called ATR-FAS. Due to imaging differences caused by various attack types, traditional FAS methods based on single binary classification network may result in excessive intra-class distance of spoof faces, leading to a challenge of decision boundary learning. Therefore, we employed multiple networks to reconstruct multi-frame depth maps as auxiliary supervision, and each network experts in one type of attack. A dual gate module (DGM) consisting of a type gate and a frame-attention gate is introduced, which perform attack type recognition and multi-frame attention generation, respectively. The outputs of DGM are utilized as weight to mix the result of multiple expert networks. The multi-experts mixture enables ATR-FAS to generate spoof-differentiated depth maps, and stably detects spoof faces without being affected by different types of PAs. Moreover, we design a differential normalization procedure to convert original flash frames into differential frames. This simple but effective processing enhances the details in flash frames, aiding in the generation of depth maps. To verify the effectiveness of our framework, we collected a large-scale dataset containing 12,660 live and spoof videos with diverse PAs under dynamic flash from the smartphone screen. Extensive experiments illustrate that the proposed ATR-FAS significantly outperforms existing state-of-the-art methods. The code and dataset will be available at https://github.com/Chaochao-Lin/ATR-FAS.

cs.CV↗
arXiv2023

IWEK: An Interpretable What-If Estimator for Database Knobs

The knobs of modern database management systems have significant impact on the performance of the systems. With the development of cloud databases, an estimation service for knobs is urgently needed to improve the performance of database. Unfortunately, few attentions have been paid to estimate the performance of certain knob configurations. To fill this gap, we propose IWEK, an interpretable & transferable what-if estimator for database knobs. To achieve interpretable estimation, we propose linear estimator based on the random forest for database knobs for the explicit and trustable evaluation results. Due to its interpretability, our estimator capture the direct relationships between knob configuration and its performance, to guarantee the high availability of database. We design a two-stage transfer algorithm to leverage historical experiences to efficiently build the knob estimator for new scenarios. Due to its lightweight design, our method can largely reduce the overhead of collecting training data and could achieve cold start knob estimation for new scenarios. Extensive experiments on YCSB and TPCC show that our method performs well in interpretable and transferable knob estimation with limited training data. Further, our method could achieve efficient estimator transfer with only 10 samples in TPCC and YSCB.

cs.DB↗