arXiv subjects

Guihai Chen

Publications and source records attributed to Guihai Chen.

At least 37 records · Page 2Linked to original sources

arXiv2026

CoDec: Prefix-Shared Decoding Kernel for LLMs

Prefix-sharing among multiple prompts presents opportunities to combine the operations of the shared prefix, while attention computation in the decode stage, which becomes a critical bottleneck with increasing context lengths, is a memory-intensive process requiring heavy memory access on the key-value (KV) cache of the prefixes. Therefore, in this paper, we explore the potential of prefix-sharing in the attention computation of the decode stage. However, the tree structure of the prefix-sharing mechanism presents significant challenges for attention computation in efficiently processing shared KV cache access patterns while managing complex dependencies and balancing irregular workloads. To address the above challenges, we propose a dedicated attention kernel to combine the memory access of shared prefixes in the decoding stage, namely CoDec. CoDec delivers two key innovations: a novel shared-prefix attention kernel that optimizes memory hierarchy and exploits both intra-block and inter-block parallelism, and a comprehensive workload balancing mechanism that efficiently estimates cost, divides tasks, and schedules execution. Experimental results show that CoDec achieves an average $1.9\times$ speedup and $120.9\times$ memory access reduction compared to the state-of-the-art FlashDecoding kernel regarding attention computation in the decode stage and $3.8\times$ end-to-end time per output token compared to the vLLM.

cs.LG↗

arXiv2026

Optimizing Feature Extraction for On-device Model Inference with User Behavior Sequences

Machine learning models are widely integrated into modern mobile apps to analyze user behaviors and deliver personalized services. Ensuring low-latency on-device model execution is critical for maintaining high-quality user experiences. While prior research has primarily focused on accelerating model inference with given input features, we identify an overlooked bottleneck in real-world on-device model execution pipelines: extracting input features from raw application logs. In this work, we explore a new direction of feature extraction optimization by analyzing and eliminating redundant extraction operations across different model features and consecutive model inferences. We then introduce AutoFeature, an automated feature extraction engine designed to accelerate on-device feature extraction process without compromising model inference accuracy. AutoFeature comprises three core designs: (1) graph abstraction to formulate the extraction workflows of different input features as one directed acyclic graph, (2) graph optimization to identify and fuse redundant operation nodes across different features within the graph; (3) efficient caching to minimize operations on overlapping raw data between consecutive model inferences. We implement a system prototype of AutoFeature and integrate it into five industrial mobile services spanning search, video and e-commerce domains. Online evaluations show that AutoFeature reduces end-to-end on-device model execution latency by 1.33x-3.93x during daytime and 1.43x-4.53x at night.

cs.LG↗

arXiv2026

AnyPro: Preference-Preserving Anycast Optimization based on Strategic AS-Path Prepending

Operating large-scale anycast networks is challenging because client-to-site mappings often misalign with operator's expectation due to opaque inter-domain routing. We present AnyPro, the first system to unlock the full potential of AS-path prepending (ASPP), efficiently deriving globally optimal configurations to steer clients toward performance-optimal sites at scale. AnyPro first employs an efficient polling mechanism to identify all clients sensitive to ASPP. By analyzing the routing changes during the process, the system derives a set of ASPP constraints that guide client traffic toward the desired sites. We then formulate the anycast optimization problem as a constraint-based program and compute optimal ASPP configurations. Extensive evaluation on a global testbed with 20 PoPs demonstrates the effectiveness of AnyPro: it reduces the 90th percentile latency by 37.7% compared to baseline configurations without ASPP. Furthermore, we show that AnyPro can be integrated with PoP-level anycast optimization techniques to achieve additional performance gains.

cs.NI↗

arXiv2026

AdaFuse: Accelerating Dynamic Adapter Inference via Token-Level Pre-Gating and Fused Kernel Optimization

The integration of dynamic, sparse structures like Mixture-of-Experts (MoE) with parameter-efficient adapters (e.g., LoRA) is a powerful technique for enhancing Large Language Models (LLMs). However, this architectural enhancement comes at a steep cost: despite minimal increases in computational load, the inference latency often skyrockets, leading to decoding speeds slowing by over 2.5 times. Through a fine-grained performance analysis, we pinpoint the primary bottleneck not in the computation itself, but in the severe overhead from fragmented, sequential CUDA kernel launches required for conventional dynamic routing. To address this challenge, we introduce AdaFuse, a framework built on a tight co-design between the algorithm and the underlying hardware system to enable efficient dynamic adapter execution. Departing from conventional layer-wise or block-wise routing, AdaFuse employs a token-level pre-gating strategy, which makes a single, global routing decision for all adapter layers before a token is processed. This "decide-once, apply-everywhere" approach effectively staticizes the execution path for each token, creating an opportunity for holistic optimization. We capitalize on this by developing a custom CUDA kernel that performs a fused switching operation, merging the parameters of all selected LoRA adapters into the backbone model in a single, efficient pass. Experimental results on popular open-source LLMs show that AdaFuse achieves accuracy on par with state-of-the-art dynamic adapters while drastically cutting decoding latency by a factor of over 2.4x, thereby bridging the gap between model capability and inference efficiency.

cs.AI↗

arXiv2026

DVD-Quant: Data-free Video Diffusion Transformers Quantization

Diffusion Transformers (DiTs) have emerged as the state-of-the-art architecture for video generation, yet their computational and memory demands hinder practical deployment. While post-training quantization (PTQ) presents a promising approach to accelerate Video DiT models, existing methods suffer from two critical limitations: (1) dependence on computation-heavy and inflexible calibration procedures, and (2) considerable performance deterioration after quantization. To address these challenges, we propose DVD-Quant, a novel Data-free quantization framework for Video DiTs. Our approach integrates three key innovations: (1) Bounded-init Grid Refinement (BGR) and (2) Auto-scaling Rotated Quantization (ARQ) for calibration data-free quantization error reduction, as well as (3) $δ$-Guided Bit Switching ($δ$-GBS) for adaptive bit-width allocation. Extensive experiments across multiple video generation benchmarks demonstrate that DVD-Quant achieves an approximately 2$\times$ speedup over full-precision baselines on advanced DiT models while maintaining visual fidelity. Notably, DVD-Quant is the first to enable W4A4 PTQ for Video DiTs without compromising video quality. Code and models will be available at https://github.com/lhxcs/DVD-Quant.

cs.CV↗

arXiv2026

You Need an Encoder for Native Position-Independent Caching

The Key-Value (KV) cache of Large Language Models (LLMs) is prefix-based, making it highly inefficient for processing contexts retrieved in arbitrary order. Position-Independent Caching (PIC) has been proposed to enable KV reuse without positional constraints; however, existing approaches often incur substantial accuracy degradation, limiting their practical adoption. To address this issue, we propose native PIC by reintroducing the encoder to prevalent decoder-only LLMs and explicitly training it to support PIC. We further develop COMB, a PIC-aware caching system that integrates seamlessly with existing inference frameworks. Experimental results show that COMB reduces Time-to-First-Token (TTFT) by 51-94% and increases throughput by 3$\times$ with comparable accuracy. Furthermore, the quality improvement when using DeepSeek-V2-Lite-Chat demonstrates the applicability of COMB to other types of decoder-only LLMs. Our code is available at https://github.com/shijuzhao/Comb.

cs.LG↗

arXiv2026

Guiding the Recommender: Information-Aware Auto-Bidding for Content Promotion

Modern content platforms offer paid promotion to mitigate cold start by allocating exposure via auctions. Our empirical analysis reveals a counterintuitive flaw in this paradigm: while promotion rescues low-to-medium quality content, it can harm high-quality content by forcing exposure to suboptimal audiences, polluting engagement signals and downgrading future recommendation. We recast content promotion as a dual-objective optimization that balances short-term value acquisition with long-term model improvement. To make this tractable at bid time in content promotion, we introduce a decomposable surrogate objective, gradient coverage, and establish its formal connection to Fisher Information and optimal experimental design. We design a two-stage auto-bidding algorithm based on Lagrange duality that dynamically paces budget through a shadow price and optimizes impression-level bids using per-impression marginal utilities. To address missing labels at bid time, we propose a confidence-gated gradient heuristic, paired with a zeroth-order variant for black-box models that reliably estimates learning signals in real time. We provide theoretical guarantees, proving monotone submodularity of the composite objective, sublinear regret in online auction, and budget feasibility. Extensive offline experiments on synthetic and real-world datasets validate the framework: it outperforms baselines, achieves superior final AUC/LogLoss, adheres closely to budget targets, and remains effective when gradients are approximated zeroth-order. These results show that strategic, information-aware promotion can improve long-term model performance and organic outcomes beyond naive impression-maximization strategies.

cs.GT↗

arXiv2026

FlexSpec: Frozen Drafts Meet Evolving Targets in Edge-Cloud Collaborative LLM Speculative Decoding

Deploying large language models (LLMs) in mobile and edge computing environments is constrained by limited on-device resources, scarce wireless bandwidth, and frequent model evolution. Although edge-cloud collaborative inference with speculative decoding (SD) can reduce end-to-end latency by executing a lightweight draft model at the edge and verifying it with a cloud-side target model, existing frameworks fundamentally rely on tight coupling between the two models. Consequently, repeated model synchronization introduces excessive communication overhead, increasing end-to-end latency, and ultimately limiting the scalability of SD in edge environments. To address these limitations, we propose FlexSpec, a communication-efficient collaborative inference framework tailored for evolving edge-cloud systems. The core design of FlexSpec is a shared-backbone architecture that allows a single and static edge-side draft model to remain compatible with a large family of evolving cloud-side target models. By decoupling edge deployment from cloud-side model updates, FlexSpec eliminates the need for edge-side retraining or repeated model downloads, substantially reducing communication and maintenance costs. Furthermore, to accommodate time-varying wireless conditions and heterogeneous device constraints, we develop a channel-aware adaptive speculation mechanism that dynamically adjusts the speculative draft length based on real-time channel state information and device energy budgets. Extensive experiments demonstrate that FlexSpec achieves superior performance compared to conventional SD approaches in terms of inference efficiency.

cs.DC↗

arXiv2025

B2LoRa: Boosting LoRa Transmission for Satellite-IoT Systems with Blind Coherent Combining

With the rapid growth of Low Earth Orbit (LEO) satellite networks, satellite-IoT systems using the LoRa technique have been increasingly deployed to provide widespread Internet services to low-power and low-cost ground devices. However, the long transmission distance and adverse environments from IoT satellites to ground devices pose a huge challenge to link reliability, as evidenced by the measurement results based on our real-world setup. In this paper, we propose a blind coherent combining design named B2LoRa to boost LoRa transmission performance. The intuition behind B2LoRa is to leverage the repeated broadcasting mechanism inherent in satellite-IoT systems to achieve coherent combining under the low-power and low-cost constraints, where each re-transmission at different times is regarded as the same packet transmitted from different antenna elements within an antenna array. Then, the problem is translated into aligning these packets at a fine granularity despite the time, frequency, and phase offsets between packets in the case of frequent packet loss. To overcome this challenge, we present three designs - joint packet sniffing, frequency shift alignment, and phase drift mitigation to deal with ultra-low SNRs and Doppler shifts featured in satellite-IoT systems, respectively. Finally, experiment results based on our real-world deployments demonstrate the high efficiency of B2LoRa.

cs.NI↗

arXiv2025

QuantFace: Efficient Quantization for Face Restoration

Diffusion models have been achieving remarkable performance in face restoration. However, the heavy computations hamper the widespread adoption of these models. In this work, we propose QuantFace, a novel low-bit quantization framework for face restoration models, where the full-precision (i.e., 32-bit) weights and activations are quantized to 4~6-bit. We first analyze the data distribution within activations and find that it is highly variant. To preserve the original data information, we employ rotation-scaling channel balancing. Furthermore, we propose Quantization-Distillation Low-Rank Adaptation (QD-LoRA), which jointly optimizes for quantization and distillation performance. Finally, we propose an adaptive bit-width allocation strategy. We formulate such a strategy as an integer programming problem that combines quantization error and perceptual metrics to find a satisfactory resource allocation. Extensive experiments on the synthetic and real-world datasets demonstrate the effectiveness of QuantFace under 6-bit and 4-bit. QuantFace achieves significant advantages over recent leading low-bit quantization methods for face restoration. The code is available at https://github.com/jiatongli2024/QuantFace.

cs.CV↗

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

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

Bi-Level Decision-Focused Causal Learning for Large-Scale Marketing Optimization: Bridging Observational and Experimental Data

Online Internet platforms require sophisticated marketing strategies to optimize user retention and platform revenue -- a classical resource allocation problem. Traditional solutions adopt a two-stage pipeline: machine learning (ML) for predicting individual treatment effects to marketing actions, followed by operations research (OR) optimization for decision-making. This paradigm presents two fundamental technical challenges. First, the prediction-decision misalignment: Conventional ML methods focus solely on prediction accuracy without considering downstream optimization objectives, leading to improved predictive metrics that fail to translate to better decisions. Second, the bias-variance dilemma: Observational data suffers from multiple biases (e.g., selection bias, position bias), while experimental data (e.g., randomized controlled trials), though unbiased, is typically scarce and costly -- resulting in high-variance estimates. We propose Bi-level Decision-Focused Causal Learning (Bi-DFCL) that systematically addresses these challenges. First, we develop an unbiased estimator of OR decision quality using experimental data, which guides ML model training through surrogate loss functions that bridge discrete optimization gradients. Second, we establish a bi-level optimization framework that jointly leverages observational and experimental data, solved via implicit differentiation. This novel formulation enables our unbiased OR estimator to correct learning directions from biased observational data, achieving optimal bias-variance tradeoff. Extensive evaluations on public benchmarks, industrial marketing datasets, and large-scale online A/B tests demonstrate the effectiveness of Bi-DFCL, showing statistically significant improvements over state-of-the-art. Currently, Bi-DFCL has been deployed at Meituan, one of the largest online food delivery platforms in the world.

cs.LG↗

arXiv2025

RAGRouter: Learning to Route Queries to Multiple Retrieval-Augmented Language Models

Retrieval-Augmented Generation (RAG) significantly improves the performance of Large Language Models (LLMs) on knowledge-intensive tasks. However, varying response quality across LLMs under RAG necessitates intelligent routing mechanisms, which select the most suitable model for each query from multiple retrieval-augmented LLMs via a dedicated router model. We observe that external documents dynamically affect LLMs' ability to answer queries, while existing routing methods, which rely on static parametric knowledge representations, exhibit suboptimal performance in RAG scenarios. To address this, we formally define the new retrieval-augmented LLM routing problem, incorporating the influence of retrieved documents into the routing framework. We propose RAGRouter, a RAG-aware routing design, which leverages document embeddings and RAG capability embeddings with contrastive learning to capture knowledge representation shifts and enable informed routing decisions. Extensive experiments on diverse knowledge-intensive tasks and retrieval settings, covering open and closed-source LLMs, show that RAGRouter outperforms the best individual LLM and existing routing methods. With an extended score-threshold-based mechanism, it also achieves strong performance-efficiency trade-offs under low-latency constraints. The code and data are available at https://github.com/OwwO99/RAGRouter.

cs.CL↗

arXiv2025

CORE: Reducing UI Exposure in Mobile Agents via Collaboration Between Cloud and Local LLMs

Mobile agents rely on Large Language Models (LLMs) to plan and execute tasks on smartphone user interfaces (UIs). While cloud-based LLMs achieve high task accuracy, they require uploading the full UI state at every step, exposing unnecessary and often irrelevant information. In contrast, local LLMs avoid UI uploads but suffer from limited capacity, resulting in lower task success rates. We propose $\textbf{CORE}$, a $\textbf{CO}$llaborative framework that combines the strengths of cloud and local LLMs to $\textbf{R}$educe UI $\textbf{E}$xposure, while maintaining task accuracy for mobile agents. CORE comprises three key components: (1) $\textbf{Layout-aware block partitioning}$, which groups semantically related UI elements based on the XML screen hierarchy; (2) $\textbf{Co-planning}$, where local and cloud LLMs collaboratively identify the current sub-task; and (3) $\textbf{Co-decision-making}$, where the local LLM ranks relevant UI blocks, and the cloud LLM selects specific UI elements within the top-ranked block. CORE further introduces a multi-round accumulation mechanism to mitigate local misjudgment or limited context. Experiments across diverse mobile apps and tasks show that CORE reduces UI exposure by up to 55.6% while maintaining task success rates slightly below cloud-only agents, effectively mitigating unnecessary privacy exposure to the cloud. The code is available at https://github.com/Entropy-Fighter/CORE.

cs.CL↗

arXiv2025

Optimizing Storage Overhead of User Behavior Log for ML-embedded Mobile Apps

Machine learning (ML) models are increasingly integrated into modern mobile apps to enable personalized and intelligent services. These models typically rely on rich input features derived from historical user behaviors to capture user intents. However, as ML-driven services become more prevalent, recording necessary user behavior data imposes substantial storage cost on mobile apps, leading to lower system responsiveness and more app uninstalls. To address this storage bottleneck, we present AdaLog, a lightweight and adaptive system designed to improve the storage efficiency of user behavior log in ML-embedded mobile apps, without compromising model inference accuracy or latency. We identify two key inefficiencies in current industrial practices of user behavior log: (i) redundant logging of overlapping behavior data across different features and models, and (ii) sparse storage caused by storing behaviors with heterogeneous attribute descriptions in a single log file. To solve these issues, AdaLog first formulates the elimination of feature-level redundant data as a maximum weighted matching problem in hypergraphs, and proposes a hierarchical algorithm for efficient on-device deployment. Then, AdaLog employs a virtually hashed attribute design to distribute heterogeneous behaviors into a few log files with physically dense storage. Finally, to ensure scalability to dynamic user behavior patterns, AdaLog designs an incremental update mechanism to minimize the I/O operations needed for adapting outdated behavior log. We implement a prototype of AdaLog and deploy it into popular mobile apps in collaboration with our industry partner. Evaluations on real-world user data show that AdaLog reduces behavior log size by 19% to 44% with minimal system overhead (only 2 seconds latency and 15 MB memory usage), providing a more efficient data foundation for broader adoption of on-device ML.

cs.LG↗

arXiv2025

TokenFlow: Responsive LLM Text Streaming Serving under Request Burst via Preemptive Scheduling

Real-time LLM interactions demand streamed token generations, where text tokens are progressively generated and delivered to users while balancing two objectives: responsiveness (i.e., low time-to-first-token) and steady generation (i.e.,required time-between-tokens). Standard LLM serving systems suffer from the inflexibility caused by non-preemptive request scheduling and reactive memory management, leading to poor resource utilization and low request processing parallelism under request bursts. Therefore, we present TokenFlow, a novel LLM serving system with enhanced text streaming performance via preemptive request scheduling and proactive key-value (KV) cache management. TokenFlow dynamically prioritizes requests based on real-time token buffer occupancy and token consumption rate, while actively transferring KV cache between GPU and CPU memory in the background and overlapping I/O with computation to minimize request preemption overhead. Extensive experiments on Llama3-8B and Qwen2.5-32B across multiple GPUs (RTX 4090, A6000, H200) demonstrate that TokenFlow achieves up to 82.5% higher effective throughput (accounting for actual user consumption) while reducing P99 TTFT by up to 80.2%, without degrading overall token throughput.

cs.LG↗

arXiv2025

Nova: Real-Time Agentic Vision-Language Model Serving with Adaptive Cross-Stage Parallelization

This paper presents Nova, a real-time scheduling framework for serving agentic vision-language models (VLMs) on a single GPU with balanced per-request latency and overall request process throughput. Our design begins by enabling effective pipelining across vision encode, LLM prefill, and LLM decode stages of VLMs, by exploiting their heterogeneous resource demands during execution and incorporating elastic GPU spatial partitioning among stages to maximally utilize the compute and memory resources. Building on this, we introduce a real-time scheduling algorithm that adaptively calibrates resource allocation among stages based on a Pareto-optimal analysis of the latency-throughput trade-off, allowing the system to sustain responsiveness and resource efficiency under dynamic request loads. To further alleviate GPU memory pressure, we design a lightweight weight offloading strategy for vision encoders that preserves inference efficiency with minimized memory overhead. Extensive evaluations on both synthetic and real-world agent workloads demonstrate that Nova consistently outperforms the state-of-the-art baselines, improving the maximum latency by up to 23.3%, while keeping competitive throughput.

cs.OS↗