arXiv subjects

Qiaoling Chen

Publications and source records attributed to Qiaoling Chen.

At least 19 recordsLinked to original sources

arXiv2026

PlexRL: Cluster-Level Orchestration of Serviceized LLM Execution for RLVR

Reinforcement learning with verifiable rewards (RLVR) has recently unlocked strong reasoning capabilities in large language models (LLMs), triggering rapid exploration of new algorithms and data. However, RLVR training is notoriously inefficient: long-tailed rollouts, tool-induced stalls, and asymmetric resource requirements between rollout and training introduce substantial idle time that cannot be eliminated by job-local optimizations such as synchronous pipelining, asynchronous rollout, or colocated execution. We argue that this inefficiency is structural. While idle gaps are unavoidable within individual RLVR jobs, they are largely anti-correlated across jobs and therefore exploitable at the cluster level. Leveraging this observation, we present PlexRL, a cluster-level runtime for multiplexing unified LLM services across RLVR jobs. By centrally managing model placement, state transitions, and function-level scheduling under strict affinity constraints, PlexRL time-slices LLM execution across jobs to fill otherwise idle periods without expensive model migration. Our implementation and evaluations demonstrate that PlexRL significantly improves effective cluster capacity and reduces user GPU hour cost by maximum 37.58% while preserving algorithmic flexibility and introducing minimal per-job overhead.

cs.DC↗

arXiv2026

SpecForge: A Flexible and Efficient Open-Source Training Framework for Speculative Decoding

Large language models incur high inference latency due to sequential autoregressive decoding. Speculative decoding alleviates this bottleneck by using a lightweight draft model to propose multiple tokens for batched verification. However, its adoption has been limited by the lack of high-quality draft models and scalable training infrastructure. We introduce SpecForge, an open-source, production-oriented framework for training speculative decoding models with full support for EAGLE-3. SpecForge incorporates target-draft decoupling, hybrid parallelism, optimized training kernels, and integration with production-grade inference engines, enabling up to 9.9x faster EAGLE-3 training for Qwen3-235B-A22B. In addition, we release SpecBundle, a suite of production-grade EAGLE-3 draft models trained with SpecForge for mainstream open-source LLMs. Through a systematic study of speculative decoding training recipes, SpecBundle addresses the scarcity of high-quality drafts in the community, and our draft models achieve up to 4.48x end-to-end inference speedup on SGLang, establishing SpecForge as a practical foundation for real-world speculative decoding deployment.

cs.LG↗

arXiv2026

CONCUR: High-Throughput Agentic Batch Inference of LLM via Congestion-Based Concurrency Control

Batch inference for agentic workloads stresses the GPU key-value (KV) cache in a sustained and cumulative manner, often causing severe throughput degradation well before memory capacity is exhausted. We identify this phenomenon as middle-phase thrashing, a previously under-characterized pathology in which cache efficiency collapses as long-lived agents accumulate state over time. We argue that mitigating this pathology requires moving beyond reactive, request-level cache management to proactive, agent-level admission control. Drawing inspiration from congestion control in distributed systems, we view the KV cache as a shared resource whose efficient utilization depends on feedback-driven regulation. Based on this insight, we present CONCUR, a lightweight control layer that regulates agent admission to bound aggregate cache pressure while preserving execution continuity. CONCUR adapts a cache-aware control algorithm to dynamically adjust the number of active agents using runtime cache signals. Across large models and real-world agent workloads, CONCUR prevents middle-phase thrashing and improves batch inference throughput by up to 4.09x on Qwen3-32B and 1.9x on DeepSeek-V3, while remaining compatible with existing LLM serving systems.

cs.DC↗

arXiv2025

Nex-N1: Agentic Models Trained via a Unified Ecosystem for Large-Scale Environment Construction

The evolution of Large Language Models (LLMs) from passive responders to autonomous agents necessitates a fundamental shift in learning paradigms -- from static imitation to incentive-driven decision making. However, this transition is significantly impeded by the lack of scalable infrastructure capable of constructing high-quality interaction signals for effective policy learning. To address this, we introduce a comprehensive method designed to systematically scale the diversity and complexity of interactive environments. Our method realizes this scaling by addressing three orthogonal dimensions: (1) Complexity: NexAU, a flexible agent framework that supports building complex agent hierarchies via simple configurations; (2) Diversity: NexA4A automatically generates diverse agent hierarchies from natural language to cover infinite domains; and (3) Fidelity: NexGAP bridges the simulation-reality gap by integrating dynamic real-world environment for grounded trajectories synthesis. We train Nex-N1 upon the diverse and complex interactive environments established by our infrastructure. Empirical results on benchmarks such as SWE-bench and tau2 demonstrate that Nex-N1 consistently outperforms SOTA open-source models and achieves competitive performance against frontier proprietary models on complex agentic tasks. We open-source the Nex ecosystem and model weights to facilitate further research.

cs.CL↗

arXiv2025

ReSpec: Towards Optimizing Speculative Decoding in Reinforcement Learning Systems

Adapting large language models (LLMs) via reinforcement learning (RL) is often bottlenecked by the generation stage, which can consume over 75\% of the training time. Speculative decoding (SD) accelerates autoregressive generation in serving systems, but its behavior under RL training remains largely unexplored. We identify three critical gaps that hinder the naive integration of SD into RL systems: diminishing speedups at large batch sizes, drafter staleness under continual actor updates, and drafter-induced policy degradation. To address these gaps, we present ReSpec, a system that adapts SD to RL through three complementary mechanisms: dynamically tuning SD configurations, evolving the drafter via knowledge distillation, and weighting updates by rollout rewards. On Qwen models (3B--14B), ReSpec achieves up to 4.5x speedup while preserving reward convergence and training stability, providing a practical solution for efficient RL-based LLM adaptation.

cs.LG↗

arXiv2025

Expert-as-a-Service: Towards Efficient, Scalable, and Robust Large-scale MoE Serving

Mixture-of-Experts (MoE) models challenge serving infrastructures with dynamic, sparse expert utilization, causing instability on conventional systems designed for dense architectures. We propose EaaS, a novel serving system to enable efficient, scalable, and robust MoE deployment. Our system disaggregates MoE modules into independent, stateless services. This design enables fine-grained resource scaling and provides inherent fault tolerance by decoupling compute units. The architecture is powered by a high-performance, CPU-free peer-to-peer communication library that ensures minimal overhead and high throughput. Experiments confirm EaaS's scalability and efficiency, achieving performance comparable to monolithic systems while providing robust fault tolerance and strong scalability. EaaS incurs less than a 2% throughput reduction under simulated hardware failures that would otherwise halt monolithic architectures. It further saves up to 37.5% of computing resources through dynamic fine-grained adaptation to serving traffic, demonstrating strong resilience for large-scale MoE deployment in production.

cs.DC↗

arXiv2025

SPPO:Efficient Long-sequence LLM Training via Adaptive Sequence Pipeline Parallel Offloading

In recent years, Large Language Models (LLMs) have exhibited remarkable capabilities, driving advancements in real-world applications. However, training LLMs on increasingly long input sequences imposes significant challenges due to high GPU memory and computational demands. Existing solutions face two key limitations: (1) memory reduction techniques, such as activation recomputation and CPU offloading, compromise training efficiency; (2) distributed parallelism strategies require excessive GPU resources, limiting the scalability of input sequence length. To address these gaps, we propose Adaptive Sequence Pipeline Parallel Offloading (SPPO), a novel LLM training framework that optimizes memory and computational resource efficiency for long-sequence training. SPPO introduces adaptive offloading, leveraging sequence-aware offloading, and two-level activation management to reduce GPU memory consumption without degrading the training efficiency. Additionally, SPPO develops an adaptive pipeline scheduling approach with a heuristic solver and multiplexed sequence partitioning to improve computational resource efficiency. Experimental results demonstrate that SPPO achieves up to 3.38x throughput improvement over Megatron-LM and DeepSpeed, realizing efficient training of a 7B LLM with sequence lengths of up to 4M tokens on only 128 A100 GPUs.

cs.DC↗

arXiv2024

LoongTrain: Efficient Training of Long-Sequence LLMs with Head-Context Parallelism

Efficiently training LLMs with long sequences is important yet challenged by the massive computation and memory requirements. Sequence parallelism has been proposed to tackle these problems, but existing methods suffer from scalability or efficiency issues. We propose LoongTrain, a novel system to efficiently train LLMs with long sequences at scale. The core of LoongTrain is the 2D-Attention mechanism, which combines both head-parallel and context-parallel techniques to break the scalability constraints while maintaining efficiency. We introduce Double-Ring-Attention and analyze the performance of device placement strategies to further speed up training. We implement LoongTrain with the hybrid ZeRO and Selective Checkpoint++ techniques. Experiment results show that LoongTrain outperforms state-of-the-art baselines, i.e., DeepSpeed-Ulysses and Megatron Context Parallelism, in both end-to-end training speed and scalability, and improves Model FLOPs Utilization (MFU) by up to 2.88x.

cs.DC↗

arXiv2024

Characterization of Large Language Model Development in the Datacenter

Large Language Models (LLMs) have presented impressive performance across several transformative tasks. However, it is non-trivial to efficiently utilize large-scale cluster resources to develop LLMs, often riddled with numerous challenges such as frequent hardware failures, intricate parallelization strategies, and imbalanced resource utilization. In this paper, we present an in-depth characterization study of a six-month LLM development workload trace collected from our GPU datacenter Acme. Specifically, we investigate discrepancies between LLMs and prior task-specific Deep Learning (DL) workloads, explore resource utilization patterns, and identify the impact of various job failures. Our analysis summarizes hurdles we encountered and uncovers potential opportunities to optimize systems tailored for LLMs. Furthermore, we introduce our system efforts: (1) fault-tolerant pretraining, which enhances fault tolerance through LLM-involved failure diagnosis and automatic recovery. (2) decoupled scheduling for evaluation, which achieves timely performance feedback via trial decomposition and scheduling optimization.

cs.DC↗

arXiv2024

AMSP: Reducing Communication Overhead of ZeRO for Efficient LLM Training

Training large language models (LLMs) encounters challenges in GPU memory consumption due to the high memory requirements of model states. The widely used Zero Redundancy Optimizer (ZeRO) addresses this issue through strategic sharding but introduces communication challenges at scale. To tackle this problem, we propose AMSP, a system designed to optimize ZeRO for scalable LLM training. AMSP incorporates three flexible sharding strategies: Full-Replica, Full-Sharding, and Partial-Sharding, and allows each component within the model states (Parameters, Gradients, Optimizer States) to independently choose a sharding strategy as well as the device mesh. We conduct a thorough analysis of communication costs, formulating an optimization problem to discover the optimal sharding strategy. Additionally, AMSP optimizes distributed LLM training by efficiently overlapping communication with computation. Evaluations demonstrate up to 52\% Model FLOPs Utilization (MFU) when training the LLaMA-based model on 1024 GPUs, resulting in a 1.56 times improvement in training throughput compared to newly proposed systems like MiCS and ZeRO++.

cs.DC↗

arXiv2024

InternEvo: Efficient Long-sequence Large Language Model Training via Hybrid Parallelism and Redundant Sharding

Large language models (LLMs) with long sequences begin to power more and more fundamentally new applications we use every day. Existing methods for long-sequence LLM training are neither efficient nor compatible with commonly-used training algorithms such as FlashAttention. We design InternEvo to address these issues. InternEvo decouples all of the sharding dimensions into a new hierarchical space, and systematically analyzes the memory and communication cost of LLM training. Then, it generates an effective hybrid parallelism strategy. We design a new selective overlap mechanism to mitigate the communication overhead introduced by the hybrid parallelism. We also implement memory management techniques to reduce GPU memory fragmentation. Evaluation results show that InternEvo generates parallelization strategies that match or outperform existing methods in model FLOPs utilization.

cs.DC↗

arXiv2023

Proof of Proposition 3.1 in the paper titled "Backstepping control of a class of space-time-varying linear parabolic PDEs via time invariant kernel functions''

We provide a detailed proof of Proposition 3.1 in the paper titled ``Backstepping control of a class of space-time-varying linear parabolic PDEs via time invariant kernel functions''. In the paper titled ``Backstepping control of a class of space-time-varying linear parabolic PDEs via time invariant kernel functions'', we addressed the problem of exponential stabilization and continuous dependence of solutions on initial data in different norms for a class of $1$-D linear parabolic PDEs with space-time-varying coefficients under backstepping boundary control. In order to stabilize the system without involving a Gevrey-like condition or the event-triggered scheme, a boundary feedback controller was designed via a time invariant kernel function. By using the approximative Lyapunov method, the exponential stability of the closed-loop system was established in the spatial $L^{p}$-norm and $W^{1,p}$-norm, respectively, whenever $p\in [1, +\infty]$. It was also shown that the solution to the considered system depends continuously on the spatial $L^{p}$-norm and $W^{1,p}$-norm, respectively, of the initial data.

math.AP↗

arXiv2022

Exponential stabilization and continuous dependence of solutions on initial data in different norms for space-time-varying linear parabolic PDEs

For an arbitrary parameter $p\in [1,+\infty]$, we consider the problem of exponential stabilization in the spatial $L^{p}$-norm, and $W^{1,p}$-norm, respectively, for a class of anti-stable linear parabolic PDEs with space-time-varying coefficients in the absence of a Gevrey-like condition, which is often imposed on time-varying coefficients of PDEs and used to guarantee the existence of smooth (w.r.t. the time variable) kernel functions in the literature. Then, based on the obtained exponential stabilities, we show that the solution of the considered system depends continuously on the $L^{p}$-norm, and $W^{1,p}$-norm, respectively, of the initial data. In order to obtain time-independent (and thus sufficiently smooth) kernel functions without a Gevrey-like condition and deal with singularities arising in the case of $p\in[1,2)$, we apply a combinatorial method, i.e., the combination of backstepping and approximation of Lyapunov functionals (ALFs), to stabilize the considered system and establish the continuous dependence of solutions on initial data in different norms.

math.OC↗

arXiv2022

Feature Transformation for Cross-domain Few-shot Remote Sensing Scene Classification

Effectively classifying remote sensing scenes is still a challenge due to the increasing spatial resolution of remote imaging and large variances between remote sensing images. Existing research has greatly improved the performance of remote sensing scene classification (RSSC). However, these methods are not applicable to cross-domain few-shot problems where target domain is with very limited training samples available and has a different data distribution from source domain. To improve the model's applicability, we propose the feature-wise transformation module (FTM) in this paper. FTM transfers the feature distribution learned on source domain to that of target domain by a very simple affine operation with negligible additional parameters. Moreover, FTM can be effectively learned on target domain in the case of few training data available and is agnostic to specific network structures. Experiments on RSSC and land-cover mapping tasks verified its capability to handle cross-domain few-shot problems. By comparison with directly finetuning, FTM achieves better performance and possesses better transferability and fine-grained discriminability. \textit{Code will be publicly available.}

cs.CV↗

arXiv2021

A two-species competition model with mixed dispersal and free boundaries in time-periodic environment

This paper is concerned with a Lotka-Volterra type competition model with free boundaries in time-periodic environment. One species is assumed to adopt nonlocal dispersal and the other one adopts mixed dispersal, which is a combination of both random dispersal and nonlocal dispersal. We show that this free boundary problem with more general growth functions admits a unique solution defined for all time. A spreading-vanishing dichotomy is obtained and criteria for spreading and vanishing are provided. Moreover, under the weak competition condition we provide the long-time asymptotic behavior of solution when spreading occurs.

math.AP↗

arXiv2019

Existence and uniqueness of global weak solutions to a generalized Camassa-Holm equation

This paper is concerned with the existence and uniqueness of global weak solutions to a generalized Camassa-Holm equation on real line. By introducing some new variables, the equation is transformed into two different semi-linear systems. Then the existence and uniqueness of global weak solutions to the original equation are obtained from that of the two semi-linear systems, respectively.

math-ph↗

arXiv2015

A reaction-diffusion-advection competition model with two free boundaries in heterogeneous time-periodic environment

In this paper, we study the dynamics of a two-species competition model with two different free boundaries in heterogeneous time-periodic environment, where the two species adopt a combination of random movement and advection upward or downward along the resource gradient. We show that the dynamics of this model can be classified into four cases, which forms a spreading-vanishing quartering. The notion of the minimal habitat size for spreading is introduced to determine if species can always spread. Rough estimates of the asymptotic spreading speed of free boundaries and the long time behavior of solutions are also established when spreading occurs. Furthermore, some sufficient conditions for spreading and vanishing are provided.

math.AP↗

arXiv2015

The diffusive competition problem with a free boundary in heterogeneous time-periodic environment

In this paper, we consider the diffusive competition problem with a free boundary and sign-changing intrinsic growth rate in heterogeneous time-periodic environment, consisting of an invasive species with density $u$ and a native species with density $v$. We assume that $v$ undergoes diffusion and growth in $R^{N}$ , and $u$ exists initially in a ball $B_{h_0}(0)$, but invades into the environment with spreading front $\{r = h(t)\}$. The effect of the dispersal rate $d_1$, the initial occupying habitat $h_0$, the initial density $u_0$ of invasive species $u$, and the parameter $μ$ (see (1.3)) on the dynamics of this free boundary problem are studied. A spreading-vanishing dichotomy is obtained and some sufficient conditions for the invasive species spreading and vanishing are provided. Moreover, when spreading of $u$ happens, some rough estimates of the spreading speed are also given.

math.AP↗