arXiv subjects

Jianwen Luo

Publications and source records attributed to Jianwen Luo.

At least 19 recordsLinked to original sources

arXiv2026

Atria Dawn: The Dawn of Agentic Superintelligence

As AI agents become participants in the development of their successors, they reshape both the production of intelligence and the role of human researchers. We introduce Atria Dawn Preview, a foundation agentic language model designed for scientific research and engineering workflows, with the goal of expanding the frontier of agent productivity in the real world. This model is trained via a Verifiable Experience Pipeline that connects tool-mediated interactions to executable environments and externally verified outcomes. Across 16 benchmarks spanning real-world research, engineering, and digital work, Atria Dawn Preview is competitive with frontier agents and achieves the highest reported score on five of them. Beyond standalone performance, we examine the real research-and-development process behind this model as a case study of human--AI collaboration, analyzing 769 task records from 56 participants together with agent logs. When asked to evaluate completed tasks under comparable conditions, participants rated about one-third of completed AI-assisted tasks as infeasible without AI. More strikingly, agents frequently propose methods and implement revisions, while humans retain most final decisions and guide exploration through judgment and feedback. These observations indicate a shift from task-level execution to project-level partnership, with human effort concentrating on what is worth pursuing and how evidence should guide research. Progress toward more autonomous AI research must therefore advance both the capacity for discovery and the capacity for meaningful human oversight, preserving accountable human authority over the risks and direction of continued development.

cs.AI↗

arXiv2026

RMPL: Relation-aware Multi-task Progressive Learning with Stage-wise Training for Multimedia Event Extraction

Multimedia Event Extraction (MEE) aims to identify events and their arguments from documents that contain both text and images. It requires grounding event semantics across different modalities. Progress in MEE is limited by the lack of annotated training data. M2E2 is the only established benchmark, but it provides annotations only for evaluation. This makes direct supervised training impractical. Existing methods mainly rely on cross-modal alignment or inference-time prompting with Vision--Language Models (VLMs). These approaches do not explicitly learn structured event representations and often produce weak argument grounding in multimodal settings. To address these limitations, we propose RMPL, a Relation-aware Multi-task Progressive Learning framework for MEE under low-resource conditions. RMPL incorporates heterogeneous supervision from unimodal event extraction and multimedia relation extraction with stage-wise training. The model is first trained with a unified schema to learn shared event-centric representations across modalities. It is then fine-tuned for event mention identification and argument role extraction using mixed textual and visual data. Experiments on the M2E2 benchmark with multiple VLMs show consistent improvements across different modality settings.

cs.CL↗

arXiv2025

DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle

Real-world enterprise data intelligence workflows encompass data engineering that turns raw sources into analytical-ready tables and data analysis that convert those tables into decision-oriented insights. We introduce DAComp, a benchmark of 210 tasks that mirrors these complex workflows. Data engineering (DE) tasks require repository-level engineering on industrial schemas, including designing and building multi-stage SQL pipelines from scratch and evolving existing systems under evolving requirements. Data analysis (DA) tasks pose open-ended business problems that demand strategic planning, exploratory analysis through iterative coding, interpretation of intermediate results, and the synthesis of actionable recommendations. Engineering tasks are scored through execution-based, multi-metric evaluation. Open-ended tasks are assessed by a reliable, experimentally validated LLM-judge, which is guided by hierarchical, meticulously crafted rubrics. Our experiments reveal that even state-of-the-art agents falter on DAComp. Performance on DE tasks is particularly low, with success rates under 20%, exposing a critical bottleneck in holistic pipeline orchestration, not merely code generation. Scores on DA tasks also average below 40%, highlighting profound deficiencies in open-ended reasoning and demonstrating that engineering and analysis are distinct capabilities. By clearly diagnosing these limitations, DAComp provides a rigorous and realistic testbed to drive the development of truly capable autonomous data agents for enterprise settings. Our data and code are available at https://da-comp.github.io

cs.CL↗

arXiv2025

Teaching Your Models to Understand Code via Focal Preference Alignment

Preference learning extends the performance of Code LLMs beyond traditional supervised fine-tuning by leveraging relative quality comparisons. In existing approaches, a set of n candidate solutions is evaluated based on test case success rates, with the candidate demonstrating a higher pass rate being labeled as positive and its counterpart with a lower pass rate as negative. However, because this approach aligns entire failing code blocks rather than pinpointing specific errors, it lacks the granularity necessary to capture meaningful error-correction relationships. As a result, the model is unable to learn more informative error-correction patterns. To address these issues, we propose Target-DPO, a new preference alignment framework that mimics human iterative debugging to refine Code LLMs. Target-DPO explicitly locates error regions and aligns the corresponding tokens via a tailored DPO algorithm. To facilitate it, we introduce the CodeFlow dataset, where samples are iteratively refined until passing tests, with modifications capturing error corrections. Extensive experiments show that a diverse suite of Code LLMs equipped with Target-DPO achieves significant performance gains in code generation and improves on challenging tasks like BigCodeBench. In-depth analysis reveals that Target-DPO yields fewer errors. Code, model and datasets are in: https://github.com/JieWu02/Target-DPO.

cs.CL↗

arXiv2025

Physics-informed Neural Network Predictive Control for Quadruped Locomotion

This study introduces a unified control framework that addresses the challenge of precise quadruped locomotion with unknown payloads, named as online payload identification-based physics-informed neural network predictive control (OPI-PINNPC). By integrating online payload identification with physics-informed neural networks (PINNs), our approach embeds identified mass parameters directly into the neural network's loss function, ensuring physical consistency while adapting to changing load conditions. The physics-constrained neural representation serves as an efficient surrogate model within our nonlinear model predictive controller, enabling real-time optimization despite the complex dynamics of legged locomotion. Experimental validation on our quadruped robot platform demonstrates 35% improvement in position and orientation tracking accuracy across diverse payload conditions (25-100 kg), with substantially faster convergence compared to previous adaptive control methods. Our framework provides a adaptive solution for maintaining locomotion performance under variable payload conditions without sacrificing computational efficiency.

cs.RO↗

arXiv2025

GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks

Large Language Models (LLMs) have shown great promise in tool-making, yet existing frameworks often struggle to efficiently construct reliable toolsets and are limited to single-task settings. To address these challenges, we propose GATE (Graph-based Adaptive Tool Evolution), an adaptive framework that dynamically constructs and evolves a hierarchical graph of reusable tools across multiple scenarios. We evaluate GATE on open-ended tasks (Minecraft), agent-based tasks (TextCraft, DABench), and code generation tasks (MATH, Date, TabMWP). Our results show that GATE achieves up to 4.3x faster milestone completion in Minecraft compared to the previous SOTA, and provides an average improvement of 9.23% over existing tool-making methods in code generation tasks and 10.03% in agent tasks. GATE demonstrates the power of adaptive evolution, balancing tool quantity, complexity, and functionality while maintaining high efficiency. Code and data are available at \url{https://github.com/ayanami2003/GATE}.

cs.CL↗

arXiv2024

DA-Code: Agent Data Science Code Generation Benchmark for Large Language Models

We introduce DA-Code, a code generation benchmark specifically designed to assess LLMs on agent-based data science tasks. This benchmark features three core elements: First, the tasks within DA-Code are inherently challenging, setting them apart from traditional code generation tasks and demanding advanced coding skills in grounding and planning. Second, examples in DA-Code are all based on real and diverse data, covering a wide range of complex data wrangling and analytics tasks. Third, to solve the tasks, the models must utilize complex data science programming languages, to perform intricate data processing and derive the answers. We set up the benchmark in a controllable and executable environment that aligns with real-world data analysis scenarios and is scalable. The annotators meticulously design the evaluation suite to ensure the accuracy and robustness of the evaluation. We develop the DA-Agent baseline. Experiments show that although the baseline performs better than other existing frameworks, using the current best LLMs achieves only 30.5% accuracy, leaving ample room for improvement. We release our benchmark at https://da-code-bench.github.io.

cs.CL↗

arXiv2024

Xiwu: A Basis Flexible and Learnable LLM for High Energy Physics

Large Language Models (LLMs) are undergoing a period of rapid updates and changes, with state-of-the-art (SOTA) model frequently being replaced. When applying LLMs to a specific scientific field, it's challenging to acquire unique domain knowledge while keeping the model itself advanced. To address this challenge, a sophisticated large language model system named as Xiwu has been developed, allowing you switch between the most advanced foundation models and quickly teach the model domain knowledge. In this work, we will report on the best practices for applying LLMs in the field of high-energy physics (HEP), including: a seed fission technology is proposed and some data collection and cleaning tools are developed to quickly obtain domain AI-Ready dataset; a just-in-time learning system is implemented based on the vector store technology; an on-the-fly fine-tuning system has been developed to facilitate rapid training under a specified foundation model. The results show that Xiwu can smoothly switch between foundation models such as LLaMA, Vicuna, ChatGLM and Grok-1. The trained Xiwu model is significantly outperformed the benchmark model on the HEP knowledge question-and-answering and code generation. This strategy significantly enhances the potential for growth of our model's performance, with the hope of surpassing GPT-4 as it evolves with the development of open-source models. This work provides a customized LLM for the field of HEP, while also offering references for applying LLM to other fields, the corresponding codes are available on Github.

hep-ph↗

arXiv2023

Fast Sampling generative model for Ultrasound image reconstruction

Image reconstruction from radio-frequency data is pivotal in ultrafast plane wave ultrasound imaging. Unlike the conventional delay-and-sum (DAS) technique, which relies on somewhat imprecise assumptions, deep learning-based methods perform image reconstruction by training on paired data, leading to a notable enhancement in image quality. Nevertheless, these strategies often exhibit limited generalization capabilities. Recently, denoising diffusion models have become the preferred paradigm for image reconstruction tasks. However, their reliance on an iterative sampling procedure results in prolonged generation time. In this paper, we propose a novel sampling framework that concurrently enforces data consistency of ultrasound signals and data-driven priors. By leveraging the advanced diffusion model, the generation of high-quality images is substantially expedited. Experimental evaluations on an in-vivo dataset indicate that our approach with a single plane wave surpasses DAS with spatial coherent compounding of 75 plane waves.

cs.CV↗

arXiv2023

Single PW takes a shortcut to compound PW in US imaging

Reconstruction of ultrasound (US) images from radio-frequency data can be conceptualized as a linear inverse problem. Traditional deep learning approaches, which aim to improve the quality of US images by directly learning priors, often encounter challenges in generalization. Recently, diffusion-based generative models have received significant attention within the research community due to their robust performance in image reconstruction tasks. However, a limitation of these models is their inherent low speed in generating image samples from pure Gaussian noise progressively. In this study, we exploit the inherent similarity between the US images reconstructed from a single plane wave (PW) and PW compounding PWC). We hypothesize that a single PW can take a shortcut to reach the diffusion trajectory of PWC, removing the need to begin with Gaussian noise. By employing an advanced diffusion model, we demonstrate its effectiveness in US image reconstruction, achieving a substantial reduction in sampling steps. In-vivo experimental results indicate that our approach can reduce sampling steps by 60%, while preserving comparable performance metrics with the conventional diffusion model.

cs.CV↗

arXiv2023

Cross-domain Self-supervised Framework for Photoacoustic Computed Tomography Image Reconstruction

Accurate image reconstruction is crucial for photoacoustic (PA) computed tomography (PACT). Recently, deep learning has been used to reconstruct the PA image with a supervised scheme, which requires high-quality images as ground truth labels. In practice, there are inevitable trade-offs between cost and performance since the use of more channels is an expensive strategy to access more measurements. Here, we propose a cross-domain unsupervised reconstruction (CDUR) strategy with a pure transformer model, which overcomes the lack of ground truth labels from limited PA measurements. The proposed approach exploits the equivariance of PACT to achieve high performance with a smaller number of channels. We implement a self-supervised reconstruction in a model-based form. Meanwhile, we also leverage the self-supervision to enforce the measurement and image consistency on three partitions of measured PA data, by randomly masking different channels. We find that dynamically masking a high proportion of the channels, e.g., 80%, yields nontrivial self-supervisors in both image and signal domains, which decrease the multiplicity of the pseudo solution to efficiently reconstruct the image from fewer PA measurements with minimum error of the image. Experimental results on in-vivo PACT dataset of mice demonstrate the potential of our unsupervised framework. In addition, our method shows a high performance (0.83 structural similarity index (SSIM) in the extreme sparse case with 13 channels), which is close to that of supervised scheme (0.77 SSIM with 16 channels). On top of all the advantages, our method may be deployed on different trainable models in an end-to-end manner.

eess.IV↗

arXiv2023

Score-based Generative Models for Photoacoustic Image Reconstruction with Rotation Consistency Constraints

Photoacoustic tomography (PAT) is a newly emerged imaging modality which enables both high optical contrast and acoustic depth of penetration. Reconstructing images of photoacoustic tomography from limited amount of senser data is among one of the major challenges in photoacoustic imaging. Previous works based on deep learning were trained in supervised fashion, which directly map the input partially known sensor data to the ground truth reconstructed from full field of view. Recently, score-based generative models played an increasingly significant role in generative modeling. Leveraging this probabilistic model, we proposed Rotation Consistency Constrained Score-based Generative Model (RCC-SGM), which recovers the PAT images by iterative sampling between Langevin dynamics and a constraint term utilizing the rotation consistency between the images and the measurements. Our proposed method can generalize to different measurement processes (32.29 PSNR with 16 measurements under random sampling, whereas 28.50 for supervised counterpart), while supervised methods need to train on specific inverse mappings.

cs.CV↗

arXiv2022

Hessian filter-assisted full diameter at half maximum (FDHM) segmentation and quantification method for optical-resolution photoacoustic microscopy

Optical-resolution photoacoustic microscopy has been validated as a high-resolution and high-sensitivity imaging modality for angiographic studies in the past decades. Quantitative vascular analysis reveals critical information of physiological changes, where vessel segmentation is the key step. In this work, we developed a Hessian filter-assisted, adaptive thresholding vessel segmentation algorithm. Its performance is validated by a digital phantom and in vivo images. Its capability of capturing subtle vessel changes is further tested in two longitudinal studies on vascular responses to blood pressure agents. The results are compared with the widely used Hessian filter method. In the antihypotensive case, the proposed method detected a twice larger vasoconstriction than the Hessian filter method. In the antihypertensive case, the proposed method detected a vasodilation of 18.8 %, while the Hessian filter method failed in change detection. The proposed algorithm could correct errors caused by conventional segmentation methods and improve quantitative accuracy for angiographic applications.

physics.med-ph↗

arXiv2022

Deep Learning Improves Dataset Recovery for High Frame Rate Synthetic Transmit Aperture Imaging

Synthetic transmit aperture (STA) imaging can achieve optimal lateral resolution in the full field of view, at the cost of low frame rate (FR) and low signal-to-noise ratio (SNR). In our previous studies, compressed sensing based synthetic transmit aperture (CS-STA) and minimal l2-norm least squares (LS-STA) methods were proposed to recover the complete STA dataset from fewer Hadamard-encoded plane wave (PW) transmissions. Results demonstrated that, compared with STA imaging, CS/LS-STA can maintain the high resolution of STA and improve the contrast in the deep region with increased FR. However, these methods would introduce errors to the recovered STA datasets and subsequently produce severe artifacts to the beamformed images. Recently, we discovered that the theoretical explanation for the error introduced in the LS-STA-based recovery is that LS-STA method neglects the null space component of the real STA data. To deal with this problem, we propose to train a convolutional neural network (CNN) under the null space learning framework (to estimate the missing null space component) for high-accuracy recovery of the STA dataset from fewer Hadamard-encoded PW transmissions. The mapping between the low-quality STA dataset (recovered using the LS-STA method) and the corresponding high-quality STA dataset (obtained using full Hadamard-encoded STA imaging, HE-STA) was learned from phantom and in vivo samples. The performance of the proposed CNN-STA method was compared with the LS-STA, STA, and HE-STA methods, in terms of visual quality, NRMSE, gCNR, and FWHM. The results demonstrate that the proposed method can improve the recovery accuracy of the STA datasets and therefore effectively suppress the artifacts presented in the images obtained using the LS-STA method. In addition, the proposed method can maintain the high lateral resolution of STA with fewer PW transmissions, as LS-STA does.

physics.med-ph↗

arXiv2022

Kirin: A Quadruped Robot with High Payload Carrying Capability

The quadruped robot is a versatile mobile platform with potential ability for high payload carrying. However, most of the existing quadruped robots aim at high maneuverability, highly dynamic and agile locomotion. In spite of this, payload carrying is still an indispensable ability for the quadruped robots. Design of a quadruped robot with high payload capacity is yet deeply explored. In this study, a 50 kg electrically-actuated quadruped robot, Kirin, is presented to leverage the payload carrying capability. Kirin is an characterized with prismatic quasi-direct-drive (QDD) leg. This mechanism greatly augments the payload carrying capability. This study presents several design principles for the payload-carrying-oriented quadruped robots, including the mechanical design, actuator parameters selection, and locomotion control method. The theoretical analysis implies that the lifting task tends to be a bottleneck for the existing robots with the articulated knee joints. By using prismatic QDD leg, the payload carrying capability of Kirin is enhanced greatly. To demonstrate Kirin's payload carrying capability, in preliminary experiment, up to 125 kg payload lifting in static stance and 50 kg payload carrying in dynamic trotting are tested. Whole body compliance with payload carrying is also demonstrated.

cs.RO↗

arXiv2022

An Adaptive Control Algorithm for Quadruped Locomotion with Proprioceptive Linear Legs

Quadruped robots manifest great potential to traverse rough terrains with payload. Numerous traditional control methods for legged dynamic locomotion are model-based and exhibit high sensitivity to model uncertainties and payload variations. Therefore, high-performance model parameter estimation becomes indispensable. However, the inertia parameters of payload are usually unknown and dynamically changing when the quadruped robot is deployed in versatile tasks. To address this problem, online identification of the inertia parameters and the Center of Mass (CoM) position of the payload for the quadruped robots draw an increasing interest. This study presents an adaptive controller based on the online payload identification for the high payload capacity (the ratio between payload and robot's self-weight) quadruped locomotion. We name it as Adaptive Controller for Quadruped Locomotion (ACQL), which consists of a recursive update law and a control law. ACQL estimates the external forces and torques induced by the payload online. The estimation is incorporated in inverse-dynamics-based Quadratic Programming (QP) to realize a trotting gait. As such, the tracking accuracy of the robot's CoM and orientation trajectories are improved. The proposed method, ACQL, is verified in a real quadruped robot platform. Experiments prove the estimation efficacy for the payload weighing from 20 $kg$ to 75 $kg$ and loaded at different locations of the robot's torso.

cs.RO↗

arXiv2021

A semi-automatic ultrasound image analysis system for the grading diagnosis of COVID-19 pneumonia

This paper proposes a semi-automatic system based on quantitative characterization of the specific image patterns in lung ultrasound (LUS) images, in order to assess the lung conditions of patients with COVID-19 pneumonia, as well as to differentiate between the severe / and no-severe cases. Specifically, four parameters are extracted from each LUS image, namely the thickness (TPL) and roughness (RPL) of the pleural line, and the accumulated with (AWBL) and acoustic coefficient (ACBL) of B lines. 27 patients are enrolled in this study, which are grouped into 13 moderate patients, 7 severe patients and 7 critical patients. Furthermore, the severe and critical patients are regarded as the severe cases, and the moderate patients are regarded as the non-severe cases. Biomarkers among different groups are compared. Each single biomarker and a classifier with all the biomarkers as input are utilized for the binary diagnosis of severe case and non-severe case, respectively. The classifier achieves the best classification performance among all the compared methods (area under the receiver operating characteristics curve = 0.93, sensitivity = 0.93, specificity = 0.85). The proposed image analysis system could be potentially applied to the grading and prognosis evaluation of patients with COVID-19 pneumonia.

physics.med-ph↗

arXiv2021

Partial Hadamard Encoded Synthetic Transmit Aperture for High Frame Rate Imaging with Minimal l2-Norm Least Square Method

Synthetic transmit aperture (STA) ultrasound imaging is well known for ideal focusing in the full field of view. However, it suffers from low signal-to-noise ratio (SNR) and low frame rate, because each array element must be activated individually. In our previous study, we encoded all the array elements with partial Hadamard matrix and reconstructed the complete STA dataset with compressed sensing (CS) algorithm (CS-STA). As all the elements are activated in each transmission and the number of transmissions is smaller than that of STA, this method can achieve higher SNR and higher frame rate. Its main drawback is the time-consuming CS reconstruction. In this study, we accelerate the complete STA dataset reconstruction with minimal l2-norm least square method. Thanks of the orthogonality of partial Hadamard matrix, the minimal l2-norm least square solution can be easily calculated. The proposed method is tested with simulation data and experimental phantom and in-vivo data. The results demonstrate that the proposed method achieves ~5*10^3 times faster reconstruction speed than CS algorithm. The simulation results demonstrate that the proposed method is capable of achieving the same accuracy for STA dataset reconstruction as conventional CS-STA method. The simulations, phantom and in-vivo experiments show that the proposed method is capable of improving the generalized contrast-to-noise ratio (gCNR) and SNR with maintained spatial resolution and fewer transmissions, compared with STA. In conclusion, the improved image quality and reduced computational time of LS-STA pave the way for its real-time applications in the clinics.

physics.med-ph↗