arXiv subjects

Yijia Zhang

Publications and source records attributed to Yijia Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

ROMA: a Read-Only-Memory-based Accelerator for QLoRA-based On-Device LLM

As large language models (LLMs) demonstrate powerful capabilities, deploying them on edge devices has become increasingly crucial, offering advantages in privacy and real-time interaction. QLoRA has emerged as the standard approach for on-device LLMs, leveraging quantized models to reduce memory and computational costs while utilizing LoRA for task-specific adaptability. In this work, we propose ROMA, a QLoRA accelerator with a hybrid storage architecture that uses ROM for quantized base models and SRAM for LoRA weights and KV cache. Our insight is that the quantized base model is stable and converged, making it well-suited for ROM storage. Meanwhile, LoRA modules offer the flexibility to adapt to new data without requiring updates to the base model. To further reduce the area cost of ROM, we introduce a novel B-ROM design and integrate it with the compute unit to form a fused cell for efficient use of chip resources. ROMA can effectively store both a 4-bit 3B and a 2-bit 8B LLaMA model entirely on-chip, achieving a notable generation speed exceeding 20,000 tokens/s without requiring external memory.

cs.AR↗

arXiv2026

TOM: A Ternary Read-only Memory Accelerator for LLM-powered Edge Intelligence

The deployment of Large Language Models (LLMs) for real-time intelligence on edge devices is rapidly growing. However, conventional hardware architectures face a fundamental memory wall challenge, where limited on-device memory capacity and bandwidth severely constrain the size of deployable models and their inference speed, while also limiting on-device adaptation. To address this challenge, we propose TOM, a hybrid ROM-SRAM accelerator co-designed with ternary quantization, which balances extreme density with on-device tunability. TOM exploits the synergy between ternary quantization and ROM to achieve extreme memory density and bandwidth, while preserving flexibility through a hybrid ROM-SRAM architecture designed for QLoRA-based tunability. Specifically, we introduce: (1) a sparsity-aware ROM architecture that synthesizes ternary weights as standard-cell logic, eliminating area overhead from zero-valued bits; (2) a distributed processing architecture that co-locates high-density ROM banks with flexible SRAM-based QLoRA adapters and compute units; and (3) a workload-aware dynamic power gating scheme that exploits the logic-based nature of ROM to power down inactive banks, minimizing dynamic energy consumption. TOM achieves an inference throughput of 3,306 TPS using BitNet-2B model, demonstrating its effectiveness in delivering real-time, energy-efficient edge intelligence.

cs.AR↗

arXiv2026

Radio counterparts of ultraluminous X-ray sources in nearby galaxies

Radio surveys of ultraluminous X-ray sources (ULXs) allow us to find supercritically accreting compact objects (SS 433/W50 like systems) or stripped nuclear black holes in nearby galaxies. We identified 21 such objects by crossmatching a ULX catalog with the Rapid ASKAP Continuum Survey (RACS) and Very Large Array Sky Survey (VLASS). They may have a diverse population. (i) Three have a double lobed radio structure with a compact core found in two of them, and could be quasars. (ii) Five are associated with extended radio structure and star forming regions in optical, where the radio emission is likely due to star forming activities, although the steep radio spectrum up to several GHz casts doubt on that. Two of them show X-ray variability suggesting that they are ULXs embedded in star forming regions. (iii) Thirteen are associated with an unresolved radio source, with a steep spectrum seen in eight, a flat or inverted spectrum seen in two. Those with a steep spectrum are arguably candidates for SS 433/W50 like objects, with radio emission due to optically thin synchrotron radiation in a surrounding jet/wind powered nebula. Remarkable cases include NGC 925 ULX1 and NGC 6946 ULX1, which are associated with an optical nebula. Those with a flat or inverted spectrum could be accreting black holes with a compact jet, while the black hole mass is estimated to be several $10^6 - 10^8$ $M_\odot$ based on the fundamental plane. Redshift measurements are needed to firmly determine the association with their apparent host galaxy.

astro-ph.HE↗

arXiv2025

A fast powerful X-ray transient from possible tidal disruption of a white dwarf

Stars captured by black holes (BHs) can be torn apart by strong tidal forces, producing electromagnetic flares. To date, more than 100 tidal disruption events (TDEs) have been observed, each involving invariably normal gaseous stars whose debris falls onto the BH, sustaining the flares over years. White dwarfs (WDs), which are the most prevalent compact stars and a million times denser--and therefore tougher--than gaseous stars, can only be disrupted by intermediate-mass black holes (IMBHs) of 10^2--10^5 solar masses. WD-TDEs are considered to generate more powerful and short-lived flares, but their evidence has been lacking. Here we report observations of a fast and luminous X-ray transient EP250702a detected by Einstein Probe. Its one-day-long X-ray peak as luminous as 10^(47-49) erg/s showed strong recurrent flares with hard spectra extending to several tens of MeV gamma-rays, as detected by Fermi/GBM and Konus-Wind, indicating relativistic jet emission. The jet's X-ray dropped sharply from 3 x 10^49 erg/s to around 10^44 erg/s within 20 days (10 days in the source rest frame). These characteristics are inconsistent with any known transient phenomena other than a jetted-TDE evolving over an unprecedentedly short timescale, indicating the disruption of a WD by an IMBH. At late times, a new soft component progressively dominates the X-ray spectrum, exhibiting an extreme super-Eddington luminosity, which possibly originates from an accretion disc. WD-TDEs open a new window for investigating the elusive IMBHs and their surrounding stellar environments, and they are prime sources of gravitational waves in the band of space-based interferometers.

astro-ph.HE↗

arXiv2025

A Multi-Agent LLM Framework for Multi-Domain Low-Resource In-Context NER via Knowledge Retrieval, Disambiguation and Reflective Analysis

In-context learning (ICL) with large language models (LLMs) has emerged as a promising paradigm for named entity recognition (NER) in low-resource scenarios. However, existing ICL-based NER methods suffer from three key limitations: (1) reliance on dynamic retrieval of annotated examples, which is problematic when annotated data is scarce; (2) limited generalization to unseen domains due to the LLM's insufficient internal domain knowledge; and (3) failure to incorporate external knowledge or resolve entity ambiguities. To address these challenges, we propose KDR-Agent, a novel multi-agent framework for multi-domain low-resource in-context NER that integrates Knowledge retrieval, Disambiguation, and Reflective analysis. KDR-Agent leverages natural-language type definitions and a static set of entity-level contrastive demonstrations to reduce dependency on large annotated corpora. A central planner coordinates specialized agents to (i) retrieve factual knowledge from Wikipedia for domain-specific mentions, (ii) resolve ambiguous entities via contextualized reasoning, and (iii) reflect on and correct model predictions through structured self-assessment. Experiments across ten datasets from five domains demonstrate that KDR-Agent significantly outperforms existing zero-shot and few-shot ICL baselines across multiple LLM backbones. The code and data can be found at https://github.com/MWXGOD/KDR-Agent.

cs.CL↗

arXiv2025

CommonVoice-SpeechRE and RPG-MoGe: Advancing Speech Relation Extraction with a New Dataset and Multi-Order Generative Framework

Speech Relation Extraction (SpeechRE) aims to extract relation triplets directly from speech. However, existing benchmark datasets rely heavily on synthetic data, lacking sufficient quantity and diversity of real human speech. Moreover, existing models also suffer from rigid single-order generation templates and weak semantic alignment, substantially limiting their performance. To address these challenges, we introduce CommonVoice-SpeechRE, a large-scale dataset comprising nearly 20,000 real-human speech samples from diverse speakers, establishing a new benchmark for SpeechRE research. Furthermore, we propose the Relation Prompt-Guided Multi-Order Generative Ensemble (RPG-MoGe), a novel framework that features: (1) a multi-order triplet generation ensemble strategy, leveraging data diversity through diverse element orders during both training and inference, and (2) CNN-based latent relation prediction heads that generate explicit relation prompts to guide cross-modal alignment and accurate triplet generation. Experiments show our approach outperforms state-of-the-art methods, providing both a benchmark dataset and an effective solution for real-world SpeechRE. The source code and dataset are publicly available at https://github.com/NingJinzhong/SpeechRE_RPG_MoGe.

cs.CL↗

arXiv2025

Peptide2Mol: A Diffusion Model for Generating Small Molecules as Peptide Mimics for Targeted Protein Binding

Structure-based drug design has seen significant advancements with the integration of artificial intelligence (AI), particularly in the generation of hit and lead compounds. However, most AI-driven approaches neglect the importance of endogenous protein interactions with peptides, which may result in suboptimal molecule designs. In this work, we present Peptide2Mol, an E(3)-equivariant graph neural network diffusion model that generates small molecules by referencing both the original peptide binders and their surrounding protein pocket environments. Trained on large datasets and leveraging sophisticated modeling techniques, Peptide2Mol not only achieves state-of-the-art performance in non-autoregressive generative tasks, but also produces molecules with similarity to the original peptide binder. Additionally, the model allows for molecule optimization and peptidomimetic design through a partial diffusion process. Our results highlight Peptide2Mol as an effective deep generative model for generating and optimizing bioactive small molecules from protein binding pockets.

cs.LG↗

arXiv2025

CCSS: Hardware-Accelerated RTL Simulation with Fast Combinational Logic Computing and Sequential Logic Synchronization

As transistor counts in a single chip exceed tens of billions, the complexity of RTL-level simulation and verification has grown exponentially, often extending simulation campaigns to several months. In industry practice, RTL simulation is divided into two phases: functional debug and system validation. While system validation demands high simulation speed and is typically accelerated using FPGAs, functional debug relies on rapid compilation-rendering multi-core CPUs the primary choice. However, the limited simulation speed of CPUs has become a major bottleneck. To address this challenge, we propose CCSS, a scalable multi-core RTL simulation platform that achieves both fast compilation and high simulation throughput. CCSS accelerates combinational logic computation and sequential logic synchronization through specialized architecture and compilation strategies. It employs a balanced DAG partitioning method and efficient boolean computation cores for combinational logic, and adopts a low-latency network-on-chip (NoC) design to synchronize sequential states across cores efficiently. Experimental results show that CCSS delivers up to 12.9x speedup over state-of-the-art multi-core simulators.

cs.AR↗

arXiv2025

UMSPU: Universal Multi-Size Phase Unwrapping via Mutual Self-Distillation and Adaptive Boosting Ensemble Segmenters

Spatial phase unwrapping is a key technique for extracting phase information to obtain 3D morphology and other features. Modern industrial measurement scenarios demand high precision, large image sizes, and high speed. However, conventional methods struggle with noise resistance and processing speed. Current deep learning methods are limited by the receptive field size and sparse semantic information, making them ineffective for large size images. To address this issue, we propose a mutual self-distillation (MSD) mechanism and adaptive boosting ensemble segmenters to construct a universal multi-size phase unwrapping network (UMSPU). MSD performs hierarchical attention refinement and achieves cross-layer collaborative learning through bidirectional distillation, ensuring fine-grained semantic representation across image sizes. The adaptive boosting ensemble segmenters combine weak segmenters with different receptive fields into a strong one, ensuring stable segmentation across spatial frequencies. Experimental results show that UMSPU overcomes image size limitations, achieving high precision across image sizes ranging from 256*256 to 2048*2048 (an 8 times increase). It also outperforms existing methods in speed, robustness, and generalization. Its practicality is further validated in structured light imaging and InSAR. We believe that UMSPU offers a universal solution for phase unwrapping, with broad potential for industrial applications.

cs.CV↗

arXiv2025

TKwinFormer: Top k Window Attention in Vision Transformers for Feature Matching

Local feature matching remains a challenging task, primarily due to difficulties in matching sparse keypoints and low-texture regions. The key to solving this problem lies in effectively and accurately integrating global and local information. To achieve this goal, we introduce an innovative local feature matching method called TKwinFormer. Our approach employs a multi-stage matching strategy to optimize the efficiency of information interaction. Furthermore, we propose a novel attention mechanism called Top K Window Attention, which facilitates global information interaction through window tokens prior to patch-level matching, resulting in improved matching accuracy. Additionally, we design an attention block to enhance attention between channels. Experimental results demonstrate that TKwinFormer outperforms state-of-the-art methods on various benchmarks. Code is available at: https://github.com/LiaoYun0x0/TKwinFormer.

eess.IV↗

arXiv2025

RFUAV: A Benchmark Dataset for Unmanned Aerial Vehicle Detection and Identification

In this paper, we propose RFUAV as a new benchmark dataset for radio-frequency based (RF-based) unmanned aerial vehicle (UAV) identification and address the following challenges: Firstly, many existing datasets feature a restricted variety of drone types and insufficient volumes of raw data, which fail to meet the demands of practical applications. Secondly, existing datasets often lack raw data covering a broad range of signal-to-noise ratios (SNR), or do not provide tools for transforming raw data to different SNR levels. This limitation undermines the validity of model training and evaluation. Lastly, many existing datasets do not offer open-access evaluation tools, leading to a lack of unified evaluation standards in current research within this field. RFUAV comprises approximately 1.3 TB of raw frequency data collected from 37 distinct UAVs using the Universal Software Radio Peripheral (USRP) device in real-world environments. Through in-depth analysis of the RF data in RFUAV, we define a drone feature sequence called RF drone fingerprint, which aids in distinguishing drone signals. In addition to the dataset, RFUAV provides a baseline preprocessing method and model evaluation tools. Rigorous experiments demonstrate that these preprocessing methods achieve state-of-the-art (SOTA) performance using the provided evaluation tools. The RFUAV dataset and baseline implementation are publicly available at https://github.com/kitoweeknd/RFUAV/.

cs.RO↗

arXiv2025

Cognitive-Aligned Document Selection for Retrieval-augmented Generation

Large language models (LLMs) inherently display hallucinations since the precision of generated texts cannot be guaranteed purely by the parametric knowledge they include. Although retrieval-augmented generation (RAG) systems enhance the accuracy and reliability of generative models by incorporating external documents, these retrieved documents often fail to adequately support the model's responses in practical applications. To address this issue, we propose GGatrieval (Fine-\textbf{G}rained \textbf{G}rounded \textbf{A}lignment Re\textbf{trieval} for verifiable generation), which leverages an LLM to dynamically update queries and filter high-quality, reliable retrieval documents. Specifically, we parse the user query into its syntactic components and perform fine-grained grounded alignment with the retrieved documents. For query components that cannot be individually aligned, we propose a dynamic semantic compensation mechanism that iteratively refines and rewrites the query while continuously updating the retrieval results. This iterative process continues until the retrieved documents sufficiently support the query's response. Our approach introduces a novel criterion for filtering retrieved documents, closely emulating human strategies for acquiring targeted information. This ensures that the retrieved content effectively supports and verifies the generated outputs. On the ALCE benchmark, our method significantly surpasses a wide range of baselines, achieving state-of-the-art performance.

cs.AI↗

arXiv2024

Detection of an Orphan X-ray Flare from a Blazar Candidate EP240709a with Einstein Probe

Blazars are often observed to flare across multiple wavelengths. Orphan flares from blazars have been only detected a few times, providing an opportunity to understand the structure of the jet in the accreting system. We report a remarkable orphan X-ray flare from a blazar candidate EP240709a, detected by Einstein Probe (EP) in July 2024. The multi-band spectral properties and variability support EP240709a as a high-energy peaked BL Lacertae-type object. The flux in 0.5-10 keV increases by at least 28 times to the value of low state in 2020, with non-detection of remarkable flaring in other bands during the same period. EP240709a exhibits the harder-when-brighter tendency in the X-ray band during the orphan flare, while its infrared-optical spectra are featureless. We employ one-zone and two-zone leptonic synchrotron self-Compton models to perform the spectral energy distribution fitting. Detecting this rare orphan flare shows the potential of EP in discovering peculiar activities from AGN in high-cadence X-ray sky surveys.

astro-ph.HE↗

arXiv2024

STAHGNet: Modeling Hybrid-grained Heterogenous Dependency Efficiently for Traffic Prediction

Traffic flow prediction plays a critical role in the intelligent transportation system, and it is also a challenging task because of the underlying complex Spatio-temporal patterns and heterogeneities evolving across time. However, most present works mostly concentrate on solely capturing Spatial-temporal dependency or extracting implicit similarity graphs, but the hybrid-granularity evolution is ignored in their modeling process. In this paper, we proposed a novel data-driven end-to-end framework, named Spatio-Temporal Aware Hybrid Graph Network (STAHGNet), to couple the hybrid-grained heterogeneous correlations in series simultaneously through an elaborately Hybrid Graph Attention Module (HGAT) and Coarse-granularity Temporal Graph (CTG) generator. Furthermore, an automotive feature engineering with domain knowledge and a random neighbor sampling strategy is utilized to improve efficiency and reduce computational complexity. The MAE, RMSE, and MAPE are used for evaluation metrics. Tested on four real-life datasets, our proposal outperforms eight classical baselines and four state-of-the-art (SOTA) methods (e.g., MAE 14.82 on PeMSD3; MAE 18.92 on PeMSD4). Besides, extensive experiments and visualizations verify the effectiveness of each component in STAHGNet. In terms of computational cost, STAHGNet saves at least four times the space compared to the previous SOTA models. The proposed model will be beneficial for more efficient TFP as well as intelligent transport system construction.

cs.AI↗

arXiv2024

Automating Energy-Efficient GPU Kernel Generation: A Fast Search-Based Compilation Approach

Deep Neural Networks (DNNs) have revolutionized various fields, but their deployment on GPUs often leads to significant energy consumption. Unlike existing methods for reducing GPU energy consumption, which are either hardware-inflexible or limited by workload constraints, this paper addresses the problem at the GPU kernel level. We propose a novel search-based compilation method to generate energy-efficient GPU kernels by incorporating energy efficiency into the search process. To accelerate the energy evaluation process, we develop an accurate energy cost model based on high-level kernel features. Furthermore, we introduce a dynamic updating strategy for the energy cost model, reducing the need for on-device energy measurements and accelerating the search process. Our evaluation demonstrates that the proposed approach can generate GPU kernels with up to 21.69% reduced energy consumption while maintaining low latency.

cs.PF↗

arXiv2024

Reprogramming Pretrained Target-Specific Diffusion Models for Dual-Target Drug Design

Dual-target therapeutic strategies have become a compelling approach and attracted significant attention due to various benefits, such as their potential in overcoming drug resistance in cancer therapy. Considering the tremendous success that deep generative models have achieved in structure-based drug design in recent years, we formulate dual-target drug design as a generative task and curate a novel dataset of potential target pairs based on synergistic drug combinations. We propose to design dual-target drugs with diffusion models that are trained on single-target protein-ligand complex pairs. Specifically, we align two pockets in 3D space with protein-ligand binding priors and build two complex graphs with shared ligand nodes for SE(3)-equivariant composed message passing, based on which we derive a composed drift in both 3D and categorical probability space in the generative process. Our algorithm can well transfer the knowledge gained in single-target pretraining to dual-target scenarios in a zero-shot manner. We also repurpose linker design methods as strong baselines for this task. Extensive experiments demonstrate the effectiveness of our method compared with various baselines.

cs.LG↗

arXiv2024

Swin-BERT: A Feature Fusion System designed for Speech-based Alzheimer's Dementia Detection

Speech is usually used for constructing an automatic Alzheimer's dementia (AD) detection system, as the acoustic and linguistic abilities show a decline in people living with AD at the early stages. However, speech includes not only AD-related local and global information but also other information unrelated to cognitive status, such as age and gender. In this paper, we propose a speech-based system named Swin-BERT for automatic dementia detection. For the acoustic part, the shifted windows multi-head attention that proposed to extract local and global information from images, is used for designing our acoustic-based system. To decouple the effect of age and gender on acoustic feature extraction, they are used as an extra input of the designed acoustic system. For the linguistic part, the rhythm-related information, which varies significantly between people living with and without AD, is removed while transcribing the audio recordings into transcripts. To compensate for the removed rhythm-related information, the character-level transcripts are proposed to be used as the extra input of a word-level BERT-style system. Finally, the Swin-BERT combines the acoustic features learned from our proposed acoustic-based system with our linguistic-based system. The experiments are based on the two datasets provided by the international dementia detection challenges: the ADReSS and ADReSSo. The results show that both the proposed acoustic and linguistic systems can be better or comparable with previous research on the two datasets. Superior results are achieved by the proposed Swin-BERT system on the ADReSS and ADReSSo datasets, which are 85.58\% F-score and 87.32\% F-score respectively.

eess.AS↗

arXiv2024

Diff4VS: HIV-inhibiting Molecules Generation with Classifier Guidance Diffusion for Virtual Screening

The AIDS epidemic has killed 40 million people and caused serious global problems. The identification of new HIV-inhibiting molecules is of great importance for combating the AIDS epidemic. Here, the Classifier Guidance Diffusion model and ligand-based virtual screening strategy are combined to discover potential HIV-inhibiting molecules for the first time. We call it Diff4VS. An extra classifier is trained using the HIV molecule dataset, and the gradient of the classifier is used to guide the Diffusion to generate HIV-inhibiting molecules. Experiments show that Diff4VS can generate more candidate HIV-inhibiting molecules than other methods. Inspired by ligand-based virtual screening, a new metric DrugIndex is proposed. The DrugIndex is the ratio of the proportion of candidate drug molecules in the generated molecule to the proportion of candidate drug molecules in the training set. DrugIndex provides a new evaluation method for evolving molecular generative models from a pharmaceutical perspective. Besides, we report a new phenomenon observed when using molecule generation models for virtual screening. Compared to real molecules, the generated molecules have a lower proportion that is highly similar to known drug molecules. We call it Degradation in molecule generation. Based on the data analysis, the Degradation may result from the difficulty of generating molecules with a specific structure in the generative model. Our research contributes to the application of generative models in drug design from method, metric, and phenomenon analysis.

cs.LG↗