arXiv subjects

Tao Xie

Publications and source records attributed to Tao Xie.

At least 73 records · Page 4Linked to original sources

arXiv2025

DeepServe: Serverless Large Language Model Serving at Scale

In this paper, we propose DEEPSERVE, a scalable and serverless AI platform designed to efficiently serve large language models (LLMs) at scale in cloud environments. DEEPSERVE addresses key challenges such as resource allocation, serving efficiency, and cold start latencies through four main design components. First, DEEPSERVE uses a simple serverless abstraction called the request-job-task model, which helps manage diverse AI workloads across posttraining and model-serving tasks. Second, DEEPSERVE integrates an in-house serving engine named FLOWSERVE using a microkernel-inspired design, NPU-centric execution, and SPMD-based parallelism to optimize LLM serving. Third, DEEPSERVE includes novel scheduling policies tailored for a configuration with both PD-disaggregated and PD-colocated instances. Fourth, DEEPSERVE includes optimizations such as pre-warmed pods, DRAM pre-loading, and NPU-fork, which allow DEEPSERVE to scale up to 64 instances in seconds. DEEPSERVE has been in production for over a year, operating on a large Ascend NPU cluster and providing industrystandard APIs for fine-tuning, agent serving, and model serving to our customers.

cs.DC↗

arXiv2025

LLIA -- Enabling Low-Latency Interactive Avatars: Real-Time Audio-Driven Portrait Video Generation with Diffusion Models

Diffusion-based models have gained wide adoption in the virtual human generation due to their outstanding expressiveness. However, their substantial computational requirements have constrained their deployment in real-time interactive avatar applications, where stringent speed, latency, and duration requirements are paramount. We present a novel audio-driven portrait video generation framework based on the diffusion model to address these challenges. Firstly, we propose robust variable-length video generation to reduce the minimum time required to generate the initial video clip or state transitions, which significantly enhances the user experience. Secondly, we propose a consistency model training strategy for Audio-Image-to-Video to ensure real-time performance, enabling a fast few-step generation. Model quantization and pipeline parallelism are further employed to accelerate the inference speed. To mitigate the stability loss incurred by the diffusion process and model quantization, we introduce a new inference strategy tailored for long-duration video generation. These methods ensure real-time performance and low latency while maintaining high-fidelity output. Thirdly, we incorporate class labels as a conditional input to seamlessly switch between speaking, listening, and idle states. Lastly, we design a novel mechanism for fine-grained facial expression control to exploit our model's inherent capacity. Extensive experiments demonstrate that our approach achieves low-latency, fluid, and authentic two-way communication. On an NVIDIA RTX 4090D, our model achieves a maximum of 78 FPS at a resolution of 384x384 and 45 FPS at a resolution of 512x512, with an initial video generation latency of 140 ms and 215 ms, respectively.

cs.CV↗

arXiv2025

RaaS: Reasoning-Aware Attention Sparsity for Efficient LLM Reasoning

Large Language Models (LLMs) have demonstrated strong capabilities across various domains, with recent advancements in challenging reasoning tasks such as mathematics and programming. However, solving reasoning tasks often requires an LLM to generate long sequences, incurring $O(N)$ time and memory complexities per token, where $N$ is the current sequence length. To reduce complexities, existing sparsity-based algorithms propose to retain Key-Value (KV) vectors, the intermediate representations of only the most critical tokens. However, these algorithms struggle with the "impossible trinity" of accuracy, time, and memory. For example, the state-of-the-art algorithm, Quest, achieves high accuracy with $O(L)$ time but $O(N)$ memory ($L$ is the cache budget, $L \ll N$). To address the "impossible trinity", in this paper, we identify a new attention pattern during the decode stage of reasoning tasks, where milestone tokens (analogous to lemmas in mathematical proofs) emerge, are utilized, and then become unimportant afterward. Based on this pattern, we propose a new algorithm RaaS that identifies milestone tokens and retains their KV vectors until they are no longer needed, achieving high accuracy with $O(L)$ time and $O(L)$ memory complexities.

cs.LG↗

arXiv2025

EPIC: Efficient Position-Independent Caching for Serving Large Language Models

Large Language Models (LLMs) show great capabilities in a wide range of applications, but serving them efficiently becomes increasingly challenging as requests (prompts) become more complex. Context caching improves serving performance by reusing Key-Value (KV) vectors, the intermediate representations of tokens that are repeated across requests. However, existing context caching requires exact prefix matches across requests, limiting reuse cases in settings such as few-shot learning and retrieval-augmented generation, where immutable content (e.g., documents) remains unchanged across requests but is preceded by varying prefixes. Position-Independent Caching (PIC) addresses this issue by enabling modular reuse of the KV vectors regardless of prefixes. We formalize PIC and advance prior work by introducing EPIC, a serving system incorporating our new LegoLink algorithm, which mitigates the inappropriate "attention sink" effect at every document beginning, to maintain accuracy with minimal computation. Experiments show that EPIC achieves up to 8x improvements in Time-To-First-Token (TTFT) and 7x throughput gains over existing systems, with negligible or no accuracy loss.

cs.LG↗

arXiv2025

CRAFT: Time Series Forecasting with Cross-Future Behavior Awareness

The past decades witness the significant advancements in time series forecasting (TSF) across various real-world domains, including e-commerce and disease spread prediction. However, TSF is usually constrained by the uncertainty dilemma of predicting future data with limited past observations. To settle this question, we explore the use of Cross-Future Behavior (CFB) in TSF, which occurs before the current time but takes effect in the future. We leverage CFB features and propose the CRoss-Future Behavior Awareness based Time Series Forecasting method (CRAFT). The core idea of CRAFT is to utilize the trend of cross-future behavior to mine the trend of time series data to be predicted. Specifically, to settle the sparse and partial flaws of cross-future behavior, CRAFT employs the Koopman Predictor Module to extract the key trend and the Internal Trend Mining Module to supplement the unknown area of the cross-future behavior matrix. Then, we introduce the External Trend Guide Module with a hierarchical structure to acquire more representative trends from higher levels. Finally, we apply the demand-constrained loss to calibrate the distribution deviation of prediction results. We conduct experiments on real-world dataset. Experiments on both offline large-scale dataset and online A/B test demonstrate the effectiveness of CRAFT. Our dataset and code is available at https://github.com/CRAFTinTSF/CRAFT.

cs.LG↗

arXiv2025

The First Prompt Counts the Most! An Evaluation of Large Language Models on Iterative Example-Based Code Generation

The capabilities of Large Language Models (LLMs) in code generation have been extensively studied, particularly for implementing target functionalities from natural-language descriptions. Alternatively, input-output (I/O) examples provide an accessible, unambiguous, and flexible way to describe functionalities. However, their inherent diversity, opaqueness, and incompleteness impose greater challenges for understanding and implementing the target requirements. Therefore, generating code from I/O examples (i.e., example-based code generation) provides a new perspective, allowing us to additionally evaluate LLMs' capability to infer target functionalities from limited information and to process new-form requirements. However, related research about LLMs in example-based code generation remains largely unexplored. To fill this gap, this paper presents the first comprehensive study on example-based code generation using LLMs. We adopt an iterative evaluation framework and formalize the objective of example-based code generation as two sequential sub-objectives: generating code conforming to the given examples and generating code that successfully implements the target functionalities from (iteratively) given examples. We assess six state-of-the-art LLMs using a new benchmark of 172 diverse target functionalities. The results demonstrate that when requirements are described using iterative I/O examples rather than natural language, the LLMs' score decreases by over 60%, and the vast majority (even over 95%) of successfully implemented functionalities are achieved in the first round of the iterations. Furthermore, we also find that combining I/O examples with even imprecise and fragmental natural language descriptions greatly improves LLM performance, and the selection of initial I/O examples can also influence the score, suggesting opportunities for prompt optimization.

cs.SE↗

arXiv2025

LLMigrate: Transforming "Lazy" Large Language Models into Efficient Source Code Migrators

Rewriting C code in Rust provides stronger memory safety, yet migrating large codebases such as the 32-million-line Linux kernel remains challenging. While rule-based translators (e.g., C2Rust) provide accurate yet largely unsafe Rust programs, recent Large Language Model (LLM) approaches produce more idiomatic, safe Rust programs but frequently exhibit "laziness", omitting significant portions of the target code. To address the issue, in this paper, we present LLMigrate, an LLM-based C-to-Rust translation tool that splits modules into discrete functions, translating them individually, and then reintegrating them. LLMigrate uses static analysis to retain necessary context, pairs GPT-4o (a state-of-the-art LLM) with compiler-driven translation and program-repair techniques for complex core functions, and leverages call-graph-guided translation to ensure consistent interfaces. Evaluations on three representative Linux kernel modules (math, sort, and ramfs) show that LLMigrate requires modifying less than 15\% of the target code, significantly outperforming a pure GPT-4o-based migration.

cs.PL↗

arXiv2025

EnvGS: Modeling View-Dependent Appearance with Environment Gaussian

Reconstructing complex reflections in real-world scenes from 2D images is essential for achieving photorealistic novel view synthesis. Existing methods that utilize environment maps to model reflections from distant lighting often struggle with high-frequency reflection details and fail to account for near-field reflections. In this work, we introduce EnvGS, a novel approach that employs a set of Gaussian primitives as an explicit 3D representation for capturing reflections of environments. These environment Gaussian primitives are incorporated with base Gaussian primitives to model the appearance of the whole scene. To efficiently render these environment Gaussian primitives, we developed a ray-tracing-based renderer that leverages the GPU's RT core for fast rendering. This allows us to jointly optimize our model for high-quality reconstruction while maintaining real-time rendering speeds. Results from multiple real-world and synthetic datasets demonstrate that our method produces significantly more detailed reflections, achieving the best rendering quality in real-time novel view synthesis. The code is available at https://zju3dv.github.io/envgs.

cs.CV↗

arXiv2025

Multi-view Reconstruction via SfM-guided Monocular Depth Estimation

In this paper, we present a new method for multi-view geometric reconstruction. In recent years, large vision models have rapidly developed, performing excellently across various tasks and demonstrating remarkable generalization capabilities. Some works use large vision models for monocular depth estimation, which have been applied to facilitate multi-view reconstruction tasks in an indirect manner. Due to the ambiguity of the monocular depth estimation task, the estimated depth values are usually not accurate enough, limiting their utility in aiding multi-view reconstruction. We propose to incorporate SfM information, a strong multi-view prior, into the depth estimation process, thus enhancing the quality of depth prediction and enabling their direct application in multi-view geometric reconstruction. Experimental results on public real-world datasets show that our method significantly improves the quality of depth estimation compared to previous monocular depth estimation works. Additionally, we evaluate the reconstruction quality of our approach in various types of scenes including indoor, streetscape, and aerial views, surpassing state-of-the-art MVS methods. The code and supplementary materials are available at https://zju3dv.github.io/murre/ .

cs.CV↗

arXiv2025

Ground state and magnetic transitions of the orthorhombic antiferromagnet CaCo$_2$TeO$_6$

We report the systematic synthesis, crystal structure, magnetization, and powder neutron diffraction of single crystalline and polycrystalline CaCo$_2$TeO$_6$ samples. CaCo$_2$TeO$_6$ crystallizes in an orthorhombic structure with $Pnma$ space group, featuring chains of edge-shared CoO$_6$ octahedra arranged in a honeycomb pattern. Two antiferromagnetic transitions are observed at $T$$_{N1}$ = 14.4 K and $T$$_{N2}$ = 16.2 K, corresponding to two long-range magnetic orders with propagation vectors of $\bf{k}$$_1$ = (0, 0, 0) and $\bf{k}$$_2$ = (0.125, 0, 0.25), respectively. The ground state is determined as a canted up-up-down-down zigzag spin configuration along the $c$ axis, wherein the magnetic moments of Co1 and Co2 ions are 3.4(1) and 2.1(1)$μ$$_B$, respectively. Successive spin-flop transitions appear with the increasing magnetic field applied along the easy axis ($c$ axis), accompanied by depression of the antiferromagnetic orders and enhancement of residual magnetic entropy. The field-induced spin-disordered state suggests that CaCo$_2$TeO$_6$ may be an ideal candidate for studying frustrated magnetism.

cond-mat.str-el↗

arXiv2025

Doping dependence of the magnetic ground state in the frustrated magnets Ba$_2$$M$Te$_{1-x}$W$_{x}$O$_6$ ($M$ = Mn, Co)

Theoretically, the relative change of the Heisenberg-type nearest-neighbor coupling $J_1$ and next-nearest-neighbor coupling $J_2$ in the face-centered-cubic lattice can give rise to three main antiferromagnetic orderings of type-I, type-II, and type-III. However, it is difficult to tune the $J_2/J_1$ ratio in real materials. Here, we report studies on the influence of Te$^{6+}$ and W$^{6+}$ ions replacement to the magnetic interactions and the magnetic ground states in the double-perovskite compounds Ba$_2$$M$Te$_{1-x}$W$_{x}$O$_6$ ($M$ = Mn, Co). For Ba$_2$MnTe$_{1-x}$W$_{x}$O$_6$, the W$^{6+}$ doping on Te$^{6+}$ site is successful in $0.02 \leq x \leq 0.9$ with short-range orders of the type-I ($0.02 \leq x \leq 0.08$) and type-II ($0.1 \leq x \leq 0.9$). In Ba$_2$CoTe$_{1-x}$W${_x}$O$_6$, x-ray diffraction measurements reveal two crystal structures, including the trigonal phase ($0 \leq x \leq 0.1$) and the cubic phase ($0.5 \leq x \leq 1$), between which is a miscibility gap. Two magnetic transitions are identified in the trigonal phase due to two magnetic subsystems, and the type-II magnetic order is observed in the cubic phase. Magnetic phase diagrams of Ba$_2M$Te$_{1-x}$W$_{x}$O$_6$ ($M$ = Mn, Co) are established. Our work shows that the magnetic interactions and ground states of Ba$_2$$M$Te$_{1-x}$W$_x$O$_6$ can be tuned effectively by the replacement of Te$^{6+}$ by W$^{6+}$ ions.

cond-mat.str-el↗

arXiv2025

Self-stabilized charge states in a double-decker molecular magnet on Pb(111)

Electron charging play key roles in physiochemical processes, whose intrinsic stabilization in single molecules is desirable for tailoring molecular functionality and developing molecular devices, but remains elusive on surfaces. Here, we show molecular charge states can be self-stabilized via intramolecular distortion in single bis(phthalocyaninato)terbium(III) (TbPc2) double-decker molecules, that were grown on Pb(111) substrate. Using scanning tunneling microscopy and spectroscopy, we identify fractions of TbPc2 molecules reduce to 2-fold symmetry, expressing energy-split molecular orbitals and two types of different spin states. Our first principles calculations unveil that the symmetry reduction is induced by charging-triggered Jahn-Teller distortions, which lifts the degenerate orbitals into two 2-fold symmetric orbitals. Single or double occupancy of the lower-energy orbital results in different molecular spin states. Such intramolecular distortion traps the excess electrons stably without explicit involvement of the substrate, in contrast to previously observed molecular charge states. These charged single molecule can be manipulated with the tip individually. This study offers a new avenue for tailoring the charge and spin states of molecules.

cond-mat.mes-hall↗

arXiv2025

Beyond Pass or Fail: Multi-Dimensional Benchmarking of Foundation Models for Goal-based Mobile UI Navigation

Recent advances of foundation models (FMs) have made navigating mobile applications (apps) based on high-level goal instructions within reach, with significant industrial applications such as UI testing. While existing benchmarks evaluate FM-based UI navigation using the binary pass/fail metric, they have two major limitations: they cannot reflect the complex nature of mobile UI navigation where FMs may fail for various reasons (e.g., misunderstanding instructions and failed planning), and they lack industrial relevance due to oversimplified tasks that poorly represent real-world scenarios. To address the preceding limitations, we propose Sphinx, a comprehensive benchmark for multi-dimensional evaluation of FMs in industrial settings of UI navigation. Sphinx introduces a specialized toolkit that evaluates five essential FM capabilities, providing detailed insights into failure modes such as insufficient app knowledge or planning issues. Using both popular Google Play applications and WeChat's internal UI test cases, we evaluate 8 FMs with 20 different configurations. Our results show that existing FMs universally struggle with goal-based testing tasks, primarily due to insufficient UI-specific capabilities. We summarize seven lessons learned from benchmarking FMs with Sphinx, providing clear directions for improving FM-based mobile UI navigation.

cs.SE↗

arXiv2025

Identification of the superconductivity in bilayer nickelate La$_3$Ni$_2$O$_7$ upon 100 GPa

Identification of superconductivity in the Ruddlesden-Popper phases of nickelates under high pressure remains challenging. Here, we report a comprehensive study of the crystal structure, resistance, and Meissner effect in single crystals of La$_3$Ni$_2$O$_7$ with hydrostatic pressures up to 104 GPa. X-ray diffraction measurements reveal a structural transition from the orthorhombic to a tetragonal phase above 40 GPa. Zero resistance of the superconductivity was achieved with a maximum onset $T_c^{onset}$ of 83 K at 18.0 GPa. Superconductivity is gradually suppressed until it disappears above 80 GPa, resulting in a right-triangle-like superconducting region. The direct-current magnetic susceptibility technique successfully detected the Meissner effect in La$_3$Ni$_2$O$_7$ under pressure; the maximum superconducting volume fraction is estimated to be 62.7% at 22.0 GPa. Thus, we demonstrate the bulk nature of superconductivity in the bilayer nickelate La$_3$Ni$_2$O$_7$ single crystals under high pressure. The results reveal intimate connections among the superconductivity, oxygen content, and structure in La$_3$Ni$_2$O$_7$.

cond-mat.supr-con↗

arXiv2025

Careful About What App Promotion Ads Recommend! Detecting and Explaining Malware Promotion via App Promotion Graph

In Android apps, their developers frequently place app promotion ads, namely advertisements to promote other apps. Unfortunately, the inadequate vetting of ad content allows malicious developers to exploit app promotion ads as a new distribution channel for malware. To help detect malware distributed via app promotion ads, in this paper, we propose a novel approach, named ADGPE, that synergistically integrates app user interface (UI) exploration with graph learning to automatically collect app promotion ads, detect malware promoted by these ads, and explain the promotion mechanisms employed by the detected malware. Our evaluation on 18, 627 app promotion ads demonstrates the substantial risks in the app promotion ecosystem.

cs.CR↗

arXiv2025

Data and System Perspectives of Sustainable Artificial Intelligence

Sustainable AI is a subfield of AI for concerning developing and using AI systems in ways of aiming to reduce environmental impact and achieve sustainability. Sustainable AI is increasingly important given that training of and inference with AI models such as large langrage models are consuming a large amount of computing power. In this article, we discuss current issues, opportunities and example solutions for addressing these issues, and future challenges to tackle, from the data and system perspectives, related to data acquisition, data processing, and AI model training and inference.

cs.AI↗

arXiv2025

An Infrastructure Software Perspective Toward Computation Offloading between Executable Specifications and Foundation Models

Foundation Models (FMs) have become essential components in modern software systems, excelling in tasks such as pattern recognition and unstructured data processing. However, their capabilities are complemented by the precision, verifiability, and deterministic nature of executable specifications, such as symbolic programs. This paper explores a new perspective on computation offloading, proposing a framework that strategically distributes computational tasks between FMs and executable specifications based on their respective strengths. We discuss the potential design of an infrastructure software framework to enable this offloading, focusing on key mechanisms such as task decomposition, resource allocation, and adaptive optimization. Furthermore, we identify critical technical challenges, including semantic-gap resolution, reliability, and scalability, that must be addressed to realize this approach. By leveraging the complementary strengths of FMs and symbolic programs, this perspective lays the groundwork for advancing hybrid software systems that are both efficient and reliable.

cs.SE↗

arXiv2025

DeCon: Detecting Incorrect Assertions via Postconditions Generated by a Large Language Model

Recently, given the docstring for the target problem and the target function signature, large language models (LLMs) have been used not only to generate source code, but also to generate test cases, consisting of test inputs and assertions (e.g., in the form of checking an actual output against the expected output). However, as shown by our empirical study on assertions generated by four LLMs for the HumanEval benchmark, over 62% of the generated assertions are incorrect (i.e., failed on the ground-truth problem solution). To detect incorrect assertions (given the docstring and the target function signature along with a sample of example inputs and outputs), in this paper, we propose a new approach named DeCon to effectively detect incorrect assertions via LLM-generated postconditions for the target problem (a postcondition is a predicate that must always be true just after the execution of the ground-truth problem solution). Our approach requires a small set of I/O examples (i.e., a sample of example inputs and outputs) for the target problem (e.g., the I/O examples included in the docstring for a target problem in HumanEval). We use the given I/O examples to filter out those LLM-generated postconditions that are violated by at least one given I/O example. We then use the remaining postconditions to detect incorrect assertions as those assertions that violate at least one remaining postcondition. Experimental results show that DeCon can detect averagely more than 64% (63% and 65.5% detected by GPT-3.5 and GPT-4, respectively) incorrect assertions generated by four state-of-the-art LLMs, and DeCon can also improve the effectiveness of these LLMs in code generation by 4% in terms of Pass@1. In addition, although DeCon might filter out correct assertions, the fault-finding ability of the remaining correct assertions decreases only slightly.

cs.SE↗