arXiv subjects

Chen Tian

Publications and source records attributed to Chen Tian.

At least 37 records · Page 2Linked to original sources

arXiv2026

DART: Diffusion-Inspired Speculative Decoding for Fast LLM Inference

Speculative decoding is an effective and lossless approach for accelerating LLM inference. However, existing widely adopted model-based draft designs, such as EAGLE3, improve accuracy at the cost of multi-step autoregressive inference, resulting in high drafting latency and ultimately rendering the drafting stage itself a performance bottleneck. Inspired by diffusion-based large language models (dLLMs), we propose DART, which leverages parallel generation to reduce drafting latency. DART predicts logits for multiple future masked positions in parallel within a single forward pass based on hidden states of the target model, thereby eliminating autoregressive rollouts in the draft model while preserving a lightweight design. Based on these parallel logit predictions, we further introduce an efficient tree pruning algorithm that constructs high-quality draft token trees with N-gram-enforced semantic continuity. DART substantially reduces draft-stage overhead while preserving high draft accuracy, leading to significantly improved end-to-end decoding speed. Experimental results demonstrate that DART achieves a 2.03x--3.44x wall-clock time speedup across multiple datasets, surpassing EAGLE3 by 30% on average and offering a practical speculative decoding framework. Code is released at https://github.com/fvliang/DART.

cs.CL↗

arXiv2025

Scaling Graph Chain-of-Thought Reasoning: A Multi-Agent Framework with Efficient LLM Serving

Graph Chain-of-Thought (Graph-CoT) enables large language models (LLMs) to perform step-by-step reasoning over graph-structured knowledge, but existing pipelines suffer from low accuracy, excessive token usage, high latency, and low throughput due to single-agent monolithic prompts, repeated context re-encoding, and inefficient serving execution. We present GLM, the first multi-agent Graph-CoT system co-designed with an optimized LLM serving architecture. GLM decomposes reasoning into specialized agents for classification, reasoning, action generation, and graph retrieval, enabling branching and selective context sharing to reduce prompt length and reasoning iterations while preserving reasoning quality, thereby improving accuracy and reducing overall token consumption. To scale inference, we introduce a Graph-CoT-aware LLM inference mechanism with graph-specific KV-cache management, priority-based eviction, and pipelined execution to improve serving efficiency. Experiments demonstrate that GLM improves answer accuracy by up to 38%, reduces token cost by up to 95.7%, lowers inference latency by 90.3%, and achieves up to 15.1x higher throughput compared to state-of-the-art Graph-CoT baselines, enabling efficient adoption for complex real-world reasoning at scale.

cs.LG↗

arXiv2025

Accelerating Mixture-of-Experts Inference by Hiding Offloading Latency with Speculative Decoding

Recent advancements in Mixture of Experts (MoE) models have significantly increased their parameter scale as well as model performance. Extensive offloading techniques have been proposed to address the GPU memory limitations of MoE inference. However, due to the I/O bottleneck and sparse computation of MoE models, existing offloading techniques still suffer from low hardware utilization. To fully utilize the hardware resources, we propose SpecMoEOff, which employs the speculative decoding technique to enlarge the workload of each expert. SpecMoEOff orchestrates the GPU and CPU by both theoretical and empirical roofline analysis. In addition, we develop a dedicated CPU chunked attention verification kernel to fit the speculative decoding in offloading scenarios as well as minimizing the additional overhead led by draft models. SpecMoEOff further integrates an optimizer to automatically tune the hyperparameters of speculative decoding under given hardware and workload. Experimental results show that SpecMoEOff achieves up to 2.5x decode throughput improvement over the state-of-the-art MoE offloading techniques.

cs.DC↗

arXiv2025

Revisiting Service Level Objectives and System Level Metrics in Large Language Model Serving

User experience is a critical factor Large Language Model (LLM) serving systems must consider, where service level objectives (SLOs) considering the experience of individual requests and system level metrics (SLMs) considering the overall system performance are two key performance measures. However, we observe two notable issues in existing metrics: 1) manually delaying the delivery of some tokens can improve SLOs, and 2) actively abandoning requests that do not meet SLOs can improve SLMs, both of which are counterintuitive. In this paper, we revisit SLOs and SLMs in LLM serving, and propose a new SLO that aligns with user experience. Based on the SLO, we propose a comprehensive metric framework called smooth goodput, which integrates SLOs and SLMs to reflect the nature of user experience in LLM serving. Through this unified framework, we reassess the performance of different LLM serving systems under multiple workloads. Evaluation results show that our metric framework provides a more comprehensive view of token delivery and request processing, and effectively captures the optimal point of user experience and system performance with different serving strategies.

cs.LG↗

arXiv2025

A Deep Learning System for Rapid and Accurate Warning of Acute Aortic Syndrome on Non-contrast CT in China

The accurate and timely diagnosis of acute aortic syndromes (AAS) in patients presenting with acute chest pain remains a clinical challenge. Aortic CT angiography (CTA) is the imaging protocol of choice in patients with suspected AAS. However, due to economic and workflow constraints in China, the majority of suspected patients initially undergo non-contrast CT as the initial imaging testing, and CTA is reserved for those at higher risk. In this work, we present an artificial intelligence-based warning system, iAorta, using non-contrast CT for AAS identification in China, which demonstrates remarkably high accuracy and provides clinicians with interpretable warnings. iAorta was evaluated through a comprehensive step-wise study. In the multi-center retrospective study (n = 20,750), iAorta achieved a mean area under the receiver operating curve (AUC) of 0.958 (95% CI 0.950-0.967). In the large-scale real-world study (n = 137,525), iAorta demonstrated consistently high performance across various non-contrast CT protocols, achieving a sensitivity of 0.913-0.942 and a specificity of 0.991-0.993. In the prospective comparative study (n = 13,846), iAorta demonstrated the capability to significantly shorten the time to correct diagnostic pathway. For the prospective pilot deployment that we conducted, iAorta correctly identified 21 out of 22 patients with AAS among 15,584 consecutive patients presenting with acute chest pain and under non-contrast CT protocol in the emergency department (ED) and enabled the average diagnostic time of these 21 AAS positive patients to be 102.1 (75-133) mins. Last, the iAorta can help avoid delayed or missed diagnosis of AAS in settings where non-contrast CT remains the unavoidable the initial or only imaging test in resource-constrained regions and in patients who cannot or did not receive intravenous contrast.

eess.IV↗

arXiv2025

ElasWave: An Elastic-Native System for Scalable Hybrid-Parallel Training

Large-scale LLM pretraining now runs across $10^5$--$10^6$ accelerators, making failures routine and elasticity mandatory. We posit that an elastic-native training system must jointly deliver (i) parameter consistency, (ii) low mean time to recovery (MTTR), (iii) high post-change throughput, and (iv) computation consistency. No prior system achieves all four simultaneously. To achieve these goals, we present ElasWave, which delivers per-step fault tolerance via multi-dimensional scheduling across graph, dataflow, DVFS, and RNG. ElasWave reshapes and reshards micro-batches while preserving the global batch size and gradient scale. It performs online pipeline resharding with asynchronous parameter migration and interleaves ZeRO partitions, reducing parameter recovery processes to disjoint rank-to-rank transfers. It further leverages DVFS to absorb pipeline bubbles and reshards RNG to keep computation consistency. Together, a dynamic communicator enables in-place communication group edits, while per-step in-memory snapshots support online verification and redistribution. We evaluate ElasWave on 96 NPUs and benchmark it against state-of-the-art baselines: throughput improves by $1.35\times$ over ReCycle and $1.60\times$ over TorchFT; communicator recovery completes within one second (up to $82\times/3.6\times$ faster than full/partial rebuilds); migration MTTR drops by as much as $51\%$; and convergence deviation is reduced by approximately $78\%$.

cs.DC↗

arXiv2025

Uniform Diophantine approximation on the plane for $β$-dynamical systems

In this paper, we investigate the two-dimensional uniform Diophantine approximation in $β$-dynamical systems. Let $β_i > 1(i=1,2)$ be real numbers, and let $T_{β_i}$ denote the $β_i$-transformation defined on $[0, 1]$. For each $(x, y) \in[0,1]^2$, we define the asymptotic approximation exponent $$ v_{β_1, β_2}(x, y)=\sup \left\{0 \leq v<\infty: \begin{array}{l} T_{β_1}^n x<β_1^{-n v} \\ T_{β_2}^n y<β_2^{-n v} \end{array} \text { for infinitely many } n \in \mathbb{N}\right\} \text {, } $$ and the uniform approximation exponent $$ \hat{v}_{β_1, β_2}(x, y)=\sup \left\{0 \leq \hat{v}<\infty: \forall~ N \gg 1, \exists 1 \leq n \leq N \text { such that } \begin{array}{l} T_{β_1}^n x < β_1^{-N \hat{v}} \\ T_{β_2}^n y < β_2^{-N \hat{v}} \end{array}\right\} . $$ We calculate the Hausdorff dimension of the intersection $$\left\{(x, y) \in[0,1]^2: \hat{v}_{β_1, β_2}(x, y)=\hat{v} \text { and } v_{β_1, β_2}(x, y)=v\right\}$$ for any $\hat{v}$ and $v$ satisfying $\log _{β_2}{β_1}>\frac{\hat{v}}{v}(1+v)$. As a corollary, we establish a definite formula for the Hausdorff dimension of the level set of the uniform approximation exponent.

math.DS↗

arXiv2025

Chordless Structure: A Pathway to Simple and Expressive GNNs

Researchers have proposed various methods of incorporating more structured information into the design of Graph Neural Networks (GNNs) to enhance their expressiveness. However, these methods are either computationally expensive or lacking in provable expressiveness. In this paper, we observe that the chords increase the complexity of the graph structure while contributing little useful information in many cases. In contrast, chordless structures are more efficient and effective for representing the graph. Therefore, when leveraging the information of cycles, we choose to omit the chords. Accordingly, we propose a Chordless Structure-based Graph Neural Network (CSGNN) and prove that its expressiveness is strictly more powerful than the k-hop GNN (KPGNN) with polynomial complexity. Experimental results on real-world datasets demonstrate that CSGNN outperforms existing GNNs across various graph tasks while incurring lower computational costs and achieving better performance than the GNNs of 3-WL expressiveness.

cs.LG↗

arXiv2025

Bingo: Radix-based Bias Factorization for Random Walk on Dynamic Graphs

Random walks are a primary means for extracting information from large-scale graphs. While most real-world graphs are inherently dynamic, state-of-the-art random walk engines failed to efficiently support such a critical use case. This paper takes the initiative to build a general random walk engine for dynamically changing graphs with two key principles: (i) This system should support both low-latency streaming updates and high-throughput batched updates. (ii) This system should achieve fast sampling speed while maintaining acceptable space consumption to support dynamic graph updates. Upholding both standards, we introduce Bingo, a GPU-based random walk engine for dynamically changing graphs. First, we propose a novel radix-based bias factorization algorithm to support constant time sampling complexity while supporting fast streaming updates. Second, we present a group-adaption design to reduce space consumption dramatically. Third, we incorporate GPU-aware designs to support high-throughput batched graph updates on massively parallel platforms. Together, Bingo outperforms existing efforts across various applications, settings, and datasets, achieving up to a 271.11x speedup compared to the state-of-the-art efforts.

cs.DC↗

arXiv2025

Metric results of the intersection of sets in Diophantine approximation

Let $ψ: \mathbb{R}_{>0}\rightarrow \mathbb{R}_{>0}$ be a non-increasing function. Denote by $W(ψ)$ the set of $ψ$-well-approximable points and by $E(ψ)$ the set of points $x\in[0,1]$ such that for any $0 < ε< 1$ there exist infinitely many $(p,q)\in\mathbb{Z}\times\mathbb{N} $ with $$\left(1-ε\right)ψ(q)< \left| x-\frac{p}{q}\right|< ψ(q) .$$ In this paper, we investigate the metric properties of the set $E(ψ).$ Specifically, we compute the $s$-dimensional Hausdorff measure $\mathcal{H}^s(E(ψ))$ of $E(ψ)$ for a large class of $s \in (0,1].$ Additionally, we establish that $$\dim_{\mathcal H} E(ψ_1) \times \cdots \times E(ψ_n) =\min \{ \dim_{\mathcal H} E(ψ_i)+n-1: 1\le i \le n \},$$ where $ψ_i:\mathbb{R}_{> 0}\rightarrow \mathbb{R}_{> 0} $ is a non-increasing function satisfying $ψ_i(x)=o(x^{-2}) $ for $1\le i \le n.$

math.NT↗

arXiv2025

Echo: Efficient Co-Scheduling of Hybrid Online-Offline Tasks for Large Language Model Serving

Large language models have been widely deployed in various applications, encompassing both interactive online tasks and batched offline tasks. Given the burstiness and latency sensitivity of online tasks, over-provisioning resources is common practice. This allows for the integration of latency-insensitive offline tasks during periods of low online load, enhancing resource utilization. However, strategically serving online and offline tasks through a preemption mechanism fails to fully leverage the flexibility of offline tasks and suffers from KV cache recomputation and irregular workloads. In this paper, we introduce Echo, a collaborative online-offline task serving system, including a scheduler, a KV cache manager, and estimation toolkits. The scheduler and KV cache manager work tightly to maximize the throughput of offline tasks, while the estimator further predicts execution time to ensure online task SLOs. The scheduler leverages the batch information of last iteration to reduce the search space for finding the optimal schedule. The KV cache manager sets the priority of the KV cache based on the type of tasks and the opportunity of prefix sharing to reduce the recomputation. Finally, the estimation toolkits predict the execution time, future memory consumption, and the throughput of offline tasks to guide the scheduler, KV cache manager, and the system deployer. Evaluation based on real-world workloads demonstrates that Echo can increase offline task throughput by up to $3.3\times$, while satisfying online task SLOs.

cs.DC↗

arXiv2023

ProNet: Network-level Bandwidth Sharing among Tenants in Cloud

In today's private cloud, the resource of the datacenter is shared by multiple tenants. Unlike the storage and computing resources, it's challenging to allocate bandwidth resources among tenants in private datacenter networks. State-of-the-art approaches are not effective or practical enough to meet tenants' bandwidth requirements. In this paper, we propose ProNet, a practical end-host-based solution for bandwidth sharing among tenants to meet their various demands. The key idea of ProNet is byte-counter, a mechanism to collect the bandwidth usage of tenants on end-hosts to guide the adjustment of the whole network allocation, without putting much pressure on switches. We evaluate ProNet both in our testbed and large-scale simulations. Results show that ProNet can support multiple allocation policies such as network proportionality and minimum bandwidth guarantee. Accordingly, the application-level performance is improved.

cs.NI↗

arXiv2021

TRUST: Triangle Counting Reloaded on GPUs

Triangle counting is a building block for a wide range of graph applications. Traditional wisdom suggests that i) hashing is not suitable for triangle counting, ii) edge-centric triangle counting beats vertex-centric design, and iii) communication-free and workload balanced graph partitioning is a grand challenge for triangle counting. On the contrary, we advocate that i) hashing can help the key operations for scalable triangle counting on Graphics Processing Units (GPUs), i.e., list intersection and graph partitioning, ii)vertex-centric design reduces both hash table construction cost and memory consumption, which is limited on GPUs. In addition, iii) we exploit graph and workload collaborative, and hashing-based 2D partitioning to scale vertex-centric triangle counting over 1,000 GPUswith sustained scalability. In this work, we present TRUST which performs triangle counting with the hash operation and vertex-centric mechanism at the core. To the best of our knowledge, TRUSTis the first work that achieves over one trillion Traversed Edges Per Second (TEPS) rate for triangle counting.

cs.DC↗

arXiv2020

When Machine Learning Meets Congestion Control: A Survey and Comparison

Machine learning (ML) has seen a significant surge and uptake across many diverse applications. The high flexibility, adaptability and computing capabilities it provides extends traditional approaches used in multiple fields including network operation and management. Numerous surveys have explored ML in the context of networking, such as traffic engineering, performance optimization and network security. Many ML approaches focus on clustering, classification, regression and reinforcement learning (RL). The innovation of this research and contribution of this paper lies in the detailed summary and comparison of learning-based congestion control (CC) approaches. Compared with traditional CC algorithms which are typically rule-based, capabilities to learn from historical experience are highly desirable. From the literature, it is observed that RL is a crucial trend among learning-based CC algorithms. In this paper, we explore the performance of RL-based CC algorithms and present current problems with RL-based CC algorithms. We outline challenges and trends related to learning-based CC algorithms.

cs.NI↗

arXiv2020

DHash: Enabling Dynamic and Efficient Hash Tables

Given a specified average load factor, hash tables offer the appeal of constant time lookup operations. However, hash tables could face severe hash collisions because of malicious attacks, buggy applications, or even bursts of incoming data, compromising this practical advantage. In this paper, we present DHash, a hash table that overcomes this challenge by allowing programmers to dynamically change its hash function on the fly, without affecting other concurrent operations such as lookup, insert, and delete. DHash is modular and allows programmers to select a variety of lock-free/wait-free set algorithms as the implementation of hash table buckets. With this flexibility, they can make trade-offs between the algorithm's progress guarantee, performance, and engineering efforts, and create DHash implementations that meet their requirements best. Evaluations on three types of architectures show that DHash noticeably outperforms other practical alternatives under heavy workloads. With a load factor of 20, DHash outperforms the other three most widely used hash tables by factors of 1.4-2.0, and when the load factor increases to 200, DHash is 2.3-6.2 times faster.

cs.DC↗

arXiv2019

Simultaneous Implementation Features Extraction and Recognition Using C3D Network for WiFi-based Human Activity Recognition

Human actions recognition has attracted more and more people's attention. Many technology have been developed to express human action's features, such as image, skeleton-based, and channel state information(CSI). Among them, on account of CSI's easy to be equipped and undemanding for light, and it has gained more and more attention in some special scene. However, the relationship between CSI signal and human actions is very complex, and some preliminary work must be done to make CSI features easy to understand for computer. Nowadays, many work departed CSI-based features' action dealing into two parts. One part is for features extraction and dimension reduce, and the other part is for time series problems. Some of them even omitted one of the two part work. Therefore, the accuracies of current recognition systems are far from satisfactory. In this paper, we propose a new deep learning based approach, i.e. C3D network and C3D network with attention mechanism, for human actions recognition using CSI signals. This kind of network can make feature extraction from spatial convolution and temporal convolution simultaneously, and through this network the two part of CSI-based human actions recognition mentioned above can be realized at the same time. The entire algorithm structure is simplified. The experimental results show that our proposed C3D network is able to achieve the best recognition performance for all activities when compared with some benchmark approaches.

cs.CV↗

arXiv2018

iReplayer: In-situ and Identical Record-and-Replay for Multithreaded Applications

Reproducing executions of multithreaded programs is very challenging due to many intrinsic and external non-deterministic factors. Existing RnR systems achieve significant progress in terms of performance overhead, but none targets the in-situ setting, in which replay occurs within the same process as the recording process. Also, most existing work cannot achieve identical replay, which may prevent the reproduction of some errors. This paper presents iReplayer, which aims to identically replay multithreaded programs in the original process (under the "in-situ" setting). The novel in-situ and identical replay of iReplayer makes it more likely to reproduce errors, and allows it to directly employ debugging mechanisms (e.g. watchpoints) to aid failure diagnosis. Currently, iReplayer only incurs 3% performance overhead on average, which allows it to be always enabled in the production environment. iReplayer enables a range of possibilities, and this paper presents three examples: two automatic tools for detecting buffer overflows and use-after-free bugs, and one interactive debugging tool that is integrated with GDB.

cs.OS↗

arXiv2016

Estimation of Passenger Route Choice Pattern Using Smart Card Data for Complex Metro Systems

Nowadays, metro systems play an important role in meeting the urban transportation demand in large cities. The understanding of passenger route choice is critical for public transit management. The wide deployment of Automated Fare Collection(AFC) systems opens up a new opportunity. However, only each trip's tap-in and tap-out timestamp and stations can be directly obtained from AFC system records; the train and route chosen by a passenger are unknown, which are necessary to solve our problem. While existing methods work well in some specific situations, they don't work for complicated situations. In this paper, we propose a solution that needs no additional equipment or human involvement than the AFC systems. We develop a probabilistic model that can estimate from empirical analysis how the passenger flows are dispatched to different routes and trains. We validate our approach using a large scale data set collected from the Shenzhen metro system. The measured results provide us with useful inputs when building the passenger path choice model.

cs.AI↗