arXiv subjects

Lei Xia

Publications and source records attributed to Lei Xia.

At least 19 recordsLinked to original sources

arXiv2026

Model Literacy: An Extra Summative Evaluation Factor for Visual Analytics

Understanding and enhancing visual analytics (VA) performance is important for maximizing their impact. Existing studies have successfully applied well-established summative evaluation methods from information visualization to the VA context, yet the recent emphasis on an extra data analysis/modeling stage in the VA pipeline poses an additional challenge. Inspired by the modern concept of visualization literacy, this paper examines model literacy, namely users' knowledge of the analysis model used in a VA technique, as an additional factor for VA performance. Results from a controlled study on the visual analysis of multidimensional data with two dimensionality-reduction models indicate a positive correlation between model-task accuracy and VA-task accuracy. The study involves two common dimensionality-reduction models, PCA and t-SNE. The correlation is stronger for PCA than for t-SNE in the current task design, a pattern consistent with the possibility that VA effectiveness is more closely associated with model literacy when model outputs are less directly readable from the visualization. Completion-time evidence does not show a stable efficiency gain, suggesting that differences in model intuitiveness may help explain when model knowledge shortens task completion and when it involves additional interpretive effort. The findings of this study suggest ways to further enrich VA evaluation methods and provide directions for developing more rigorous model-literacy assessment instruments.

cs.HC↗

arXiv2026

Infini Memory: Maintainable Topic Documents for Long-Term LLM Agent Memory

Long-term LLM agents need persistent memory that can track changing facts and provide relevant evidence across sessions. Existing memory systems often store observations as isolated records, summaries, or indexed fragments, which makes evidence aggregation, fact revision, and memory maintenance difficult. We propose Infini Memory, a maintainable text-based persistent memory architecture that treats agent memory as topic-structured documents. Each topic document serves as a semantic unit for collecting related evidence, preserving metadata, and revising facts over time. New observations are first staged in a buffer and periodically consolidated into coherent textual contexts. At inference time, an agentic retrieval procedure lets the LLM read memory through iterative tool calls rather than a single retrieval step. On MemoryAgentBench, Infini Memory achieves 64.7% overall score. Ablations show that topic-structured maintenance and iterative evidence inspection improve complementary aspects of long-term memory use.

cs.AI↗

arXiv2026

Visualization of Machine Learning Models through Their Spatial and Temporal Listeners

Model visualization (ModelVis) has emerged as a major research direction, yet existing taxonomies are largely organized by data or tasks, making it difficult to treat models as first-class analysis objects. We present a model-centric two-stage framework that employs abstract listeners to capture spatial and temporal model behaviors, and then connects the translated model behavior data to the classical InfoVis pipeline. To apply the framework at scale, we build a retrieval-augmented human--large language model (LLM) extraction workflow and curate a corpus of 128 VIS/VAST ModelVis papers with 331 coded figures. Our analysis shows a dominant result-centric priority on visualizing model outcomes, quantitative/nominal data type, statistical charts, and performance evaluation. Citation-weighted trends further indicate that less frequent model-mechanism-oriented studies have disproportionately high impact while are less investigated recently. Overall, the framework is a general approach for comparing existing ModelVis systems and guiding possible future designs.

cs.LG↗

arXiv2025

MOTIF-RF: Multi-template On-chip Transformer Synthesis Incorporating Frequency-domain Self-transfer Learning for RFIC Design Automation

This paper presents a systematic study on developing multi-template machine learning (ML) surrogate models and applying them to the inverse design of transformers (XFMRs) in radio-frequency integrated circuits (RFICs). Our study starts with benchmarking four widely used ML architectures, including MLP-, CNN-, UNet-, and GT-based models, using the same datasets across different XFMR topologies. To improve modeling accuracy beyond these baselines, we then propose a new frequency-domain self-transfer learning technique that exploits correlations between adjacent frequency bands, leading to around 30%-50% accuracy improvement in the S-parameters prediction. Building on these models, we further develop an inverse design framework based on the covariance matrix adaptation evolutionary strategy (CMA-ES) algorithm. This framework is validated using multiple impedance-matching tasks, all demonstrating fast convergence and trustworthy performance. These results advance the goal of AI-assisted specs-to-GDS automation for RFICs and provide RFIC designers with actionable tools for integrating AI into their workflows.

cs.LG↗

arXiv2025

Exploration of Two-Photon Exchange in proton Form Factors at BESIII

The proton and neutron make up over 99.9\% of visible matter in the universe. The internal structure of protons, governed by ElectroMagnetic Form Factors (EMFFs), has been probed in both space-like (SL) and time-like (TL) regions. The BESIII experiment has achieved the most precise TL proton FF measurements to date, providing new insights into proton structure. However, higher-order effects in EMFFs, especially the impact of Two-Photon Exchange (TPE), remain poorly understood due to experimental challenges in $e^{+}e^{-}$ annihilation. In this work, we analyze high-statistics BESIII data across a wide c.m. energy range and observe, for the first time, a significant angular asymmetry (8.6$\sigma$) in proton production. We demonstrate that this asymmetry arises from TPE, using a Born-like event selection, and extract the corresponding TPE-related EMFF properties. The observed interference between two-photon and one-photon exchanges (OPE$\otimes$TPE) is consistent with expectations at the level of the fine-structure constant relative to the OPE.

hep-ex↗

arXiv2025

Step-Audio-AQAA: a Fully End-to-End Expressive Large Audio Language Model

Large Audio-Language Models (LALMs) have significantly advanced intelligent human-computer interaction, yet their reliance on text-based outputs limits their ability to generate natural speech responses directly, hindering seamless audio interactions. To address this, we introduce Step-Audio-AQAA, a fully end-to-end LALM designed for Audio Query-Audio Answer (AQAA) tasks. The model integrates a dual-codebook audio tokenizer for linguistic and semantic feature extraction, a 130-billion-parameter backbone LLM and a neural vocoder for high-fidelity speech synthesis. Our post-training approach employs interleaved token-output of text and audio to enhance semantic coherence and combines Direct Preference Optimization (DPO) with model merge to improve performance. Evaluations on the StepEval-Audio-360 benchmark demonstrate that Step-Audio-AQAA excels especially in speech control, outperforming the state-of-art LALMs in key areas. This work contributes a promising solution for end-to-end LALMs and highlights the critical role of token-based vocoder in enhancing overall performance for AQAA tasks.

cs.SD↗

arXiv2025

Step1X-Edit: A Practical Framework for General Image Editing

In recent years, image editing models have witnessed remarkable and rapid development. The recent unveiling of cutting-edge multimodal models such as GPT-4o and Gemini2 Flash has introduced highly promising image editing capabilities. These models demonstrate an impressive aptitude for fulfilling a vast majority of user-driven editing requirements, marking a significant advancement in the field of image manipulation. However, there is still a large gap between the open-source algorithm with these closed-source models. Thus, in this paper, we aim to release a state-of-the-art image editing model, called Step1X-Edit, which can provide comparable performance against the closed-source models like GPT-4o and Gemini2 Flash. More specifically, we adopt the Multimodal LLM to process the reference image and the user's editing instruction. A latent embedding has been extracted and integrated with a diffusion image decoder to obtain the target image. To train the model, we build a data generation pipeline to produce a high-quality dataset. For evaluation, we develop the GEdit-Bench, a novel benchmark rooted in real-world user instructions. Experimental results on GEdit-Bench demonstrate that Step1X-Edit outperforms existing open-source baselines by a substantial margin and approaches the performance of leading proprietary models, thereby making significant contributions to the field of image editing.

cs.CV↗

arXiv2025

Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction

Real-time speech interaction, serving as a fundamental interface for human-machine collaboration, holds immense potential. However, current open-source models face limitations such as high costs in voice data collection, weakness in dynamic control, and limited intelligence. To address these challenges, this paper introduces Step-Audio, the first production-ready open-source solution. Key contributions include: 1) a 130B-parameter unified speech-text multi-modal model that achieves unified understanding and generation, with the Step-Audio-Chat version open-sourced; 2) a generative speech data engine that establishes an affordable voice cloning framework and produces the open-sourced lightweight Step-Audio-TTS-3B model through distillation; 3) an instruction-driven fine control system enabling dynamic adjustments across dialects, emotions, singing, and RAP; 4) an enhanced cognitive architecture augmented with tool calling and role-playing abilities to manage complex tasks effectively. Based on our new StepEval-Audio-360 evaluation benchmark, Step-Audio achieves state-of-the-art performance in human evaluations, especially in terms of instruction following. On open-source benchmarks like LLaMA Question, shows 9.3% average performance improvement, demonstrating our commitment to advancing the development of open-source multi-modal language technologies. Our code and models are available at https://github.com/stepfun-ai/Step-Audio.

cs.CL↗

arXiv2025

Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model

We present Step-Video-T2V, a state-of-the-art text-to-video pre-trained model with 30B parameters and the ability to generate videos up to 204 frames in length. A deep compression Variational Autoencoder, Video-VAE, is designed for video generation tasks, achieving 16x16 spatial and 8x temporal compression ratios, while maintaining exceptional video reconstruction quality. User prompts are encoded using two bilingual text encoders to handle both English and Chinese. A DiT with 3D full attention is trained using Flow Matching and is employed to denoise input noise into latent frames. A video-based DPO approach, Video-DPO, is applied to reduce artifacts and improve the visual quality of the generated videos. We also detail our training strategies and share key observations and insights. Step-Video-T2V's performance is evaluated on a novel video generation benchmark, Step-Video-T2V-Eval, demonstrating its state-of-the-art text-to-video quality when compared with both open-source and commercial engines. Additionally, we discuss the limitations of current diffusion-based model paradigm and outline future directions for video foundation models. We make both Step-Video-T2V and Step-Video-T2V-Eval available at https://github.com/stepfun-ai/Step-Video-T2V. The online version can be accessed from https://yuewen.cn/videos as well. Our goal is to accelerate the innovation of video foundation models and empower video content creators.

cs.CV↗

arXiv2024

Document-level Causal Relation Extraction with Knowledge-guided Binary Question Answering

As an essential task in information extraction (IE), Event-Event Causal Relation Extraction (ECRE) aims to identify and classify the causal relationships between event mentions in natural language texts. However, existing research on ECRE has highlighted two critical challenges, including the lack of document-level modeling and causal hallucinations. In this paper, we propose a Knowledge-guided binary Question Answering (KnowQA) method with event structures for ECRE, consisting of two stages: Event Structure Construction and Binary Question Answering. We conduct extensive experiments under both zero-shot and fine-tuning settings with large language models (LLMs) on the MECI and MAVEN-ERE datasets. Experimental results demonstrate the usefulness of event structures on document-level ECRE and the effectiveness of KnowQA by achieving state-of-the-art on the MECI dataset. We observe not only the effectiveness but also the high generalizability and low inconsistency of our method, particularly when with complete event structures after fine-tuning the models.

cs.CL↗

arXiv2024

Optimizing RLHF Training for Large Language Models with Stage Fusion

We present RLHFuse, an efficient training system with stage fusion for Reinforcement Learning from Human Feedback (RLHF). Due to the intrinsic nature of RLHF training, i.e., the data skewness in the generation stage and the pipeline bubbles in the training stage, existing RLHF systems suffer from low GPU utilization. RLHFuse breaks the traditional view of RLHF workflow as a composition of individual tasks, splitting each task into finer-grained subtasks, and performing stage fusion to improve GPU utilization. RLHFuse contains two key ideas. First, for generation and inference tasks, RLHFuse splits them into sample-level subtasks, enabling efficient inter-stage fusion to overlap the execution of generation and inference stages, thus mitigating the original generation bottleneck dominated by long-tailed samples. Second, for training tasks, RLHFuse breaks them into subtasks of micro-batches and performs intra-stage fusion to concurrently execute these subtasks in the training stage with a fused pipeline schedule, effectively mitigating the pipeline bubbles. The experiments show that RLHFuse increases the training throughput by up to $3.7\times$, compared to existing systems.

cs.LG↗

arXiv2024

Charm Sivers function at EicC

The Electron-Ion Collider in China (EicC) is pivotal in enhancing our knowledge of the internal structure of nucleons and nuclei, particularly through the study of transverse momentum-dependent parton distributions (TMDs). Among the leading-twist TMDs, the Sivers function is of particular interest, as it provides crucial insights into the spin and momentum structure of hadrons and plays a significant role in describing transverse single spin asymmetries (SSAs) in high-energy scatterings. In this study, we focus on the theoretical framework and phenomenological implications of the Sivers function in the context of small-x physics, where it is intricately connected to the spin-dependent QCD odderon, demonstrating that the SSA can be expressed in terms of transverse momentum-dependent factorization within the Color Glass Condensate effective theory. Furthermore, we present simulation results using PythiaeRHIC to assess the feasibility of measuring the charm quark Sivers function at EicC. The simulation outcomes suggest that EicC, with its unique kinematic coverage, offers distinct advantages for probing the Sivers function, which would provide compelling evidence for the existence of the elusive spin-dependent odderon.

hep-ph↗

arXiv2024

GeneticPrism: Multifaceted Visualization of Scientific Impact Evolutions

Understanding the evolution of scholarly impact is essential for many real-life decision-making processes in academia, such as research planning, frontier exploration, and award selection. Popular platforms like Google Scholar and Web of Science rely on numerical indicators that are too abstract to convey the context and content of scientific impact, while most existing visualization approaches on mapping science do not consider the presentation of individual scholars' impact evolution using curated self-citation data. This paper builds on our previous work and proposes an integrated pipeline to visualize a scholar's impact evolution from multiple topic facets. A novel 3D prism-shaped visual metaphor is introduced as the overview of a scholar's impact, whilst their scientific evolution on each topic is displayed in a more structured manner. Additional designs by topic chord diagram, streamgraph visualization, and inter-topic flow map, optimized by an elaborate layout algorithm, assist in perceiving the scholar's scientific evolution across topics. A new six-degree-impact glyph metaphor highlights key interdisciplinary works driving the evolution. The proposed visualization methods are evaluated through case studies analyzing the careers of prestigious Turing award laureates and a major visualization venue.

cs.DL↗

arXiv2023

TRANSOM: An Efficient Fault-Tolerant System for Training LLMs

Large language models (LLMs) with hundreds of billions or trillions of parameters, represented by chatGPT, have achieved profound impact on various fields. However, training LLMs with super-large-scale parameters requires large high-performance GPU clusters and long training periods lasting for months. Due to the inevitable hardware and software failures in large-scale clusters, maintaining uninterrupted and long-duration training is extremely challenging. As a result, A substantial amount of training time is devoted to task checkpoint saving and loading, task rescheduling and restart, and task manual anomaly checks, which greatly harms the overall training efficiency. To address these issues, we propose TRANSOM, a novel fault-tolerant LLM training system. In this work, we design three key subsystems: the training pipeline automatic fault tolerance and recovery mechanism named Transom Operator and Launcher (TOL), the training task multi-dimensional metric automatic anomaly detection system named Transom Eagle Eye (TEE), and the training checkpoint asynchronous access automatic fault tolerance and recovery technology named Transom Checkpoint Engine (TCE). Here, TOL manages the lifecycle of training tasks, while TEE is responsible for task monitoring and anomaly reporting. TEE detects training anomalies and reports them to TOL, who automatically enters the fault tolerance strategy to eliminate abnormal nodes and restart the training task. And the asynchronous checkpoint saving and loading functionality provided by TCE greatly shorten the fault tolerance overhead. The experimental results indicate that TRANSOM significantly enhances the efficiency of large-scale LLM training on clusters. Specifically, the pre-training time for GPT3-175B has been reduced by 28%, while checkpoint saving and loading performance have improved by a factor of 20.

cs.DC↗

arXiv2023

Study of charm hadronization and in-medium modification at the Electron-ion Collider in China

Charm quark production and its hadronization in ep and eA collisions at the future Electron-Ion Collider in China (EicC) will help us understand the quark/gluon fragmentation processes and the hadronization mechanisms in the nuclear medium, especially within a poorly constrained kinematic region ($x<0.1$). In this paper, we report a study on the production of charmed hadrons, $D^0$ and $Λ_c^+$, reconstructed with a dedicated GEANT4 simulation of vertex$\,\&\,$tracking detectors designed for EicC. The $Λ_c^+$/$D^0$ ratios as functions of multiplicity and $p_T$, as well as the $D^0$ double ratio are presented with projected statistical precision.

nucl-ex↗

arXiv2023

Probing gluon distributions with $D^0$ production at the EicC

The Electron-Ion Collider in China (EicC) has been proposed to study the inner structure of matter and fundamental laws of strong interactions. In this paper, we will present a conceptual design of the tracking system based on the state-of-art silicon detector and Micro-Pattern Gaseous Detector at the EicC and demonstrate that it will enable us to reconstruct charm hadron with good significance, hence study gluonic parton distribution functions in nucleons and nuclei, as well as gluon helicity distributions. The impact study using reweighting techniques shows that the impact of the EicC will be mainly in the large $x$ region. It complements similar physics programs at the Electron-Ion Collider at Brookhaven National Laboratory.

nucl-ex↗

arXiv2021

Probing gluon helicity with heavy flavor at the EIC

We propose a new measurement of the heavy flavor hadron double spin asymmetry in deep-inelastic scattering at a future Electron-Ion Collider (EIC) to constrain the polarized gluon distribution function inside the proton. Statistical projection on $D^0$ meson double spin asymmetry is calculated with an EIC central detector using an all-silicon tracker and vertexing subsystem. A first impact study was done by interpreting pseudo-data at next-to-leading order in QCD. The sensitivity of the experimental observable in constraining gluon helicity distribution in a wide range of parton momentum fraction $x$ has been investigated considering different beam energy configurations. This measurement complements the inclusive spin-dependent structure function measurement and provides an opportunity to constrain the gluon helicity distribution in the moderate $x$ region.

hep-ex↗

arXiv2021

Electron-Ion Collider in China

Lepton scattering is an established ideal tool for studying inner structure of small particles such as nucleons as well as nuclei. As a future high energy nuclear physics project, an Electron-ion collider in China (EicC) has been proposed. It will be constructed based on an upgraded heavy-ion accelerator, High Intensity heavy-ion Accelerator Facility (HIAF) which is currently under construction, together with a new electron ring. The proposed collider will provide highly polarized electrons (with a polarization of $\sim$80%) and protons (with a polarization of $\sim$70%) with variable center of mass energies from 15 to 20 GeV and the luminosity of (2-3) $\times$ 10$^{33}$ cm$^{-2}$ s$^{-1}$. Polarized deuterons and Helium-3, as well as unpolarized ion beams from Carbon to Uranium, will be also available at the EicC. The main foci of the EicC will be precision measurements of the structure of the nucleon in the sea quark region, including 3D tomography of nucleon; the partonic structure of nuclei and the parton interaction with the nuclear environment; the exotic states, especially those with heavy flavor quark contents. In addition, issues fundamental to understanding the origin of mass could be addressed by measurements of heavy quarkonia near-threshold production at the EicC. In order to achieve the above-mentioned physics goals, a hermetical detector system will be constructed with cutting-edge technologies. This document is the result of collective contributions and valuable inputs from experts across the globe. The EicC physics program complements the ongoing scientific programs at the Jefferson Laboratory and the future EIC project in the United States. The success of this project will also advance both nuclear and particle physics as well as accelerator and detector technology in China.

nucl-ex↗