arXiv subjects

Erhu Feng

Publications and source records attributed to Erhu Feng.

16 recordsLinked to original sources

arXiv2026

COPE: Composable Performance Emulation for Mixture-of-Experts Offloading Policies

As LLMs scale, mixture-of-experts (MoE) architectures increase model capacity while keeping per-token computation manageable through sparse expert activation. This sparsity also enables memory-efficient deployment by retaining most expert weights in CPU memory and transferring or executing only the routed experts on demand. Existing MoE offloading systems typically target specific model-platform configurations, yet differences in model sparsity, CPU and GPU throughput, VRAM capacity, and host-device bandwidth can change the policy ranking. Because implementing and exhaustively evaluating all candidate policies incurs substantial engineering cost, efficiently selecting the best-performing policy for each model-platform configuration remains a key challenge for practical MoE offloading. We present COPE (Composable Offloading Policy Emulator), which selects the best-performing MoE offloading policy for a target model-platform configuration before runtime integration. COPE comprises two components. Its Policy Matrix factors the design space into two independent axes, expert residency and MoE execution strategy, enabling mechanism reuse across candidate policies and exposing four new candidates. Its Virtual Heterogeneous Machine profiles combined-operator schedules on the target platform and combines these measurements with offline routing traces to estimate candidate performance. Compared with the evaluated baselines, COPE improves fastest-policy selection accuracy by 24%-100% and reduces performance-prediction error by $1.7\times$ to $6.8\times$. By identifying the preferred policy before integration, COPE further reduces policy-specific implementation effort by up to 74%.

cs.LG↗

arXiv2026

Reuse, Rewrite, or Reject: Evidence-Constrained Trace Reuse for Repeated Mobile GUI Control

Mobile GUI agents frequently perform repetitive tasks with similar interaction procedures. Reusing previous executions can reduce redundant LLM inference and execution overhead, but variations in user instructions and execution environments may invalidate previously successful actions. We systematically characterize these variations into six representative categories and observe that such changes do not necessarily invalidate entire procedures: individual transitions can remain reusable when validated against the current task context and interface state. Based on this insight, we propose graph-text templates for selective execution-trace reuse. The graph captures UI states and action transitions, while the text encodes task intent, procedural structure, and typed slots for variable content. A record-template-replay framework constructs templates from verified executions and leverages the current instruction and live UI state to reuse, adapt, or discard individual transitions. Given an execution trace, our framework reduces LLM inference token consumption by approximately 80% on average across five base agents on AndroidWorld, with most agents achieving at least 1.5-fold execution speedup over LLM-only execution. In real-device evaluations on Android and HarmonyOS, our framework achieves action reuse ratios of 70.7-79.8% while maintaining over 2-fold execution speedup in most evaluated configurations.

cs.SE↗

arXiv2026

FastEEP: Fast Intra-Instance Elastic Scaling for Expert-Parallel MoE Serving

Large-scale expert parallelism (EP) has become the standard deployment configuration for Mixture-of-Experts (MoE) models in production. Meanwhile, the highly dynamic nature of LLM workloads requires providers to continuously adjust serving capacity to match fluctuating demand. Scaling MoE serving cannot be efficiently achieved by simply adding or removing serving instances; instead, it often requires resizing the EP group within an existing instance, which we refer to as intra-instance scaling.Existing inference engines primarily rely on instance-level scaling, which destroys and reconstructs entire serving instances, incurring substantial scaling overhead. We present FastEEP (Fast Elastic Expert Parallel), a fast intra-instance scaling system for MoE serving based on the principle of Partial Reuse. FastEEP reuses existing runtime state across worker execution, GPU memory, and communication by separating reusable execution structures from the resources activated by the current EP configuration. This separation enables coordinated resource reclamation and reconfiguration while preserving costly-to-reconstruct state, allowing fast EP scaling without full instance teardown and reconstruction. Our evaluation shows that FastEEP achieves 4.5x--11.7x faster scaling than the existing vLLM-based approach across different models, enabling efficient, fine-grained scaling of MoE serving.

cs.DC↗

arXiv2026

Memorix: Managing Memory Selections for AI Agents

AI agents now handle a growing share of tasks across sessions within a persistent project. Their behavior depends not only on the model, but heavily on the persistent state they read from the environment. Existing systems assemble this memory at use time, splicing selected files into the prompt as a one-off payload. As a result, per-task assembly is costly and can still miss critical material, failures cannot be traced or replayed, and agents can read documentation written for the wrong version. Our key insight is that what an agent should read is mostly determined by its project, so the selection can be computed once, ahead of time, and reused across tasks. We propose Memorix, an agent memory system that realizes this insight. Memorix achieves this via three key mechanisms: (1) an immutable and versioned Store pins every source by identity and reproduces historical layouts cheaply, (2) a View that splits selection (Derivations) from composition (Environment), making the result reproducible and shareable, and (3) Quests and Profiles keep a delivered view refreshable and available inside the project as sources evolve. We implement Memorix and evaluate it on SWE-ContextBench and LibEvolutionEval against mainstream agent memory systems. Precomputed Environments raise resolved rate by up to 1.5X and deliver the correct library version on every version-conditioned task, while cutting amortized cost by up to 100X and build time by up to 38X.

cs.OS↗

arXiv2026

Systematic Exploration of Multi-core Architectures for Efficient LLM Serving using WaferAI-SIM

With the widespread adoption of Large Language Models (LLMs), the demand for high-performance LLM inference services continues to grow. Multi-core AI accelerators, such as Groq, Graphcore IPU, and Cerebras WSE, provide promising platforms for LLM serving, but their distributed memory systems require careful coordination between hardware configuration and serving policies. Otherwise, mismatched tensor partitioning, data placement, and memory management can substantially underutilize compute and communication resources. To address these challenges, we present WaferAI-SIM, a multi-level simulation framework that combines transaction-level simulation with an analytical performance model. WaferAI-SIM enables simulator-driven co-design of LLM serving strategies and multi-core accelerator architectures, targeting the early design stage in which emerging platforms are not yet broadly available for empirical serving studies. It captures how LLM serving policies interact with compute-core count, memory hierarchy, and interconnect topology, enabling architecture-aware exploration beyond GPU-centric assumptions. We evaluate representative LLMs across a range of chip configurations and serving scenarios. Across the evaluation, WaferAI-SIM reports 1.32$\times$--6.03$\times$ latency improvements, where the lower endpoint comes from ring-based placement at TP=16 over the placement baselines, and the upper endpoint comes from K-dimension TP over MN-dimension TP for Qwen3\_4B at TP=4 with sequence length 256. For LLM serving, our findings provide guidance for co-designing hardware architectures and serving strategies for multi-core AI accelerators across diverse LLM workloads.

cs.AR↗

arXiv2026

When Answers Stray from Questions: Hallucination Detection via Question-Answer Orthogonal Decomposition

Hallucination detection in large language models (LLMs) requires balancing accu racy, efficiency, and robustness to distribution shift. Black-box consistency methods are effective but demand repeated inference; single-pass white-box probes are effi cient yet treat answer representations in isolation, often degrading sharply under domain shift. We propose QAOD (Question-Answer Orthogonal Decomposition), a single-pass framework that projects away the question-aligned direction from the answer representation to obtain a question-orthogonal component that suppresses domain-conditioned variation. To identify informative signals, QAOD further selects layers via diversity-penalized Fisher scoring and discriminative neurons via Fisher importance. To address both in-domain detection and cross-domain generalization, we design two complementary probing strategies: pairing the or thogonal component with question context yields a joint probe that maximizes in-domain discriminability, while using the orthogonal component alone preserves domain-agnostic factuality signals for robust transfer. QAOD's joint probe achieves the best in-domain AUROC across all evaluated model-dataset pairs, while the orthogonal-only probe delivers the strongest OOD transfer, surpassing the best white-box baseline by up to 21% on BioASQ at under 25% of generation cost.

cs.LG↗

arXiv2026

SkVM: Revisiting Language VM for Skills across Heterogenous LLMs and Harnesses

LLM agents increasingly adopt skills as a reusable unit of composition. While skills are shared across diverse agent platforms, current systems treat them as raw context, causing the same skill to behave inconsistently for different agents. This fragility undermines skill portability and execution efficiency. To address this challenge, we analyze 118,000 skills and draw inspiration from traditional compiler design. We treat skills as code and LLMs as heterogeneous processors. To make portability actionable, we decompose a skill's requirements into a set of primitive capabilities, and measure how well each model-harness pair supports them. Based on these capability profiles, we propose SkVM, a compilation and runtime system designed for portable and efficient skill execution. At compile time, SkVM performs capability-based compilation, environment binding, and concurrency extraction. At runtime, SkVM applies JIT code solidification and adaptive recompilation for performance optimization. We evaluate SkVM across eight LLMs of varying scales and three agent harnesses, covering SkillsBench and representative skill tasks. Results demonstrate that SkVM significantly improves task completion rates across different models and environments while reducing token consumption by up to 40%. In terms of performance, SkVM achieves up to 3.2x speedup with enhanced parallelism, and 19-50x latency reduction through code solidification.

cs.SE↗

arXiv2026

MobiFlow: Real-World Mobile Agent Benchmarking through Trajectory Fusion

Mobile agents can autonomously complete user-assigned tasks through GUI interactions. However, existing mainstream evaluation benchmarks, such as AndroidWorld, operate by connecting to a system-level Android emulator and provide evaluation signals based on the state of system resources. In real-world mobile-agent scenarios, however, many third-party applications do not expose system-level APIs to determine whether a task has succeeded, leading to a mismatch between benchmarks and real-world usage and making it difficult to evaluate model performance accurately. To address these issues, we propose MobiFlow, an evaluation framework built on tasks drawn from arbitrary third-party applications. Using an efficient graph-construction algorithm based on multi-trajectory fusion, MobiFlow can effectively compress the state space, support dynamic interaction, and better align with real-world third-party application scenarios. MobiFlow covers 20 widely used third-party applications and comprises 240 diverse real-world tasks, with enriched evaluation metrics. Compared with AndroidWorld, MobiFlow's evaluation results show higher alignment with human assessments and can guide the training of future GUI-based models under real workloads.

cs.AI↗

arXiv2025

Beyond Training: Enabling Self-Evolution of Agents with MOBIMEM

Large Language Model (LLM) agents are increasingly deployed to automate complex workflows in mobile and desktop environments. However, current model-centric agent architectures struggle to self-evolve post-deployment: improving personalization, capability, and efficiency typically requires continuous model retraining/fine-tuning, which incurs prohibitive computational overheads and suffers from an inherent trade-off between model accuracy and inference efficiency. To enable iterative self-evolution without model retraining, we propose MOBIMEM, a memory-centric agent system. MOBIMEM first introduces three specialized memory primitives to decouple agent evolution from model weights: (1) Profile Memory uses a lightweight distance-graph (DisGraph) structure to align with user preferences, resolving the accuracy-latency trade-off in user profile retrieval; (2) Experience Memory employs multi-level templates to instantiate execution logic for new tasks, ensuring capability generalization; and (3) Action Memory records fine-grained interaction sequences, reducing the reliance on expensive model inference. Building upon this memory architecture, MOBIMEM further integrates a suite of OS-inspired services to orchestrate execution: a scheduler that coordinates parallel sub-task execution and memory operations; an agent record-and-replay (AgentRR) mechanism that enables safe and efficient action reuse; and a context-aware exception handling that ensures graceful recovery from user interruptions and runtime errors. Evaluation on AndroidWorld and top-50 apps shows that MOBIMEM achieves 83.1% profile alignment with 23.83 ms retrieval time (280x faster than GraphRAG baselines), improves task success rates by up to 50.3%, and reduces end-to-end latency by up to 9x on mobile devices.

cs.AI↗

arXiv2025

Characterizing Mobile SoC for Accelerating Heterogeneous LLM Inference

With the rapid advancement of artificial intelligence technologies such as ChatGPT, AI agents, and video generation, contemporary mobile systems have begun integrating these AI capabilities on local devices to enhance privacy and reduce response latency. To meet the computational demands of AI tasks, current mobile SoCs are equipped with diverse AI accelerators, including GPUs and Neural Processing Units (NPUs). However, there has not been a comprehensive characterization of these heterogeneous processors, and existing designs typically only leverage a single AI accelerator for LLM inference, leading to suboptimal use of computational resources and memory bandwidth. In this paper, we first summarize key performance characteristics of heterogeneous processors, SoC memory bandwidth, etc. Drawing on these observations, we propose different heterogeneous parallel mechanisms to fully exploit both GPU and NPU computational power and memory bandwidth. We further design a fast synchronization mechanism between heterogeneous processors that leverages the unified memory architecture. By employing these techniques, we present HeteroInfer, the fastest LLM inference engine in mobile devices which supports GPU-NPU heterogeneous execution. Evaluation shows that HeteroInfer delivers a 1.34x to 6.02x end-to-end speedup over state-of-the-art GPU-only and NPU-only LLM engines, while maintaining negligible interference with other applications.

cs.DC↗

arXiv2025

MobiAgent: A Systematic Framework for Customizable Mobile Agents

With the rapid advancement of Vision-Language Models (VLMs), GUI-based mobile agents have emerged as a key development direction for intelligent mobile systems. However, existing agent models continue to face significant challenges in real-world task execution, particularly in terms of accuracy and efficiency. To address these limitations, we propose MobiAgent, a comprehensive mobile agent system comprising three core components: the MobiMind-series agent models, the AgentRR acceleration framework, and the MobiFlow benchmarking suite. Furthermore, recognizing that the capabilities of current mobile agents are still limited by the availability of high-quality data, we have developed an AI-assisted agile data collection pipeline that significantly reduces the cost of manual annotation. Compared to both general-purpose LLMs and specialized GUI agent models, MobiAgent achieves state-of-the-art performance in real-world mobile scenarios.

cs.MA↗

arXiv2025

History Rhymes: Accelerating LLM Reinforcement Learning with RhymeRL

With the rapid advancement of large language models (LLMs), reinforcement learning (RL) has emerged as a pivotal methodology for enhancing the reasoning capabilities of LLMs. Unlike traditional pre-training approaches, RL encompasses multiple stages: rollout, reward, and training, which necessitates collaboration among various worker types. However, current RL systems continue to grapple with substantial GPU underutilization, due to two primary factors: (1) The rollout stage dominates the overall RL process due to test-time scaling; (2) Imbalances in rollout lengths (within the same batch) result in GPU bubbles. While prior solutions like asynchronous execution and truncation offer partial relief, they may compromise training accuracy for efficiency. Our key insight stems from a previously overlooked observation: rollout responses exhibit remarkable similarity across adjacent training epochs. Based on the insight, we introduce RhymeRL, an LLM RL system designed to accelerate RL training with two key innovations. First, to enhance rollout generation, we present HistoSpec, a speculative decoding inference engine that utilizes the similarity of historical rollout token sequences to obtain accurate drafts. Second, to tackle rollout bubbles, we introduce HistoPipe, a two-tier scheduling strategy that leverages the similarity of historical rollout distributions to balance workload among rollout workers. We have evaluated RhymeRL within a real production environment, demonstrating scalability from dozens to thousands of GPUs. Experimental results demonstrate that RhymeRL achieves a 2.6x performance improvement over existing methods, without compromising accuracy or modifying the RL paradigm.

cs.LG↗

arXiv2025

Leveraging OS-Level Primitives for Robotic Action Management

End-to-end imitation learning frameworks (e.g., VLA) are increasingly prominent in robotics, as they enable rapid task transfer by learning directly from perception to control, eliminating the need for complex hand-crafted features. However, even when employing SOTA VLA-based models, they still exhibit limited generalization capabilities and suboptimal action efficiency, due to the constraints imposed by insufficient robotic training datasets. In addition to addressing this problem using model-based approaches, we observe that robotic action slices, which consist of contiguous action steps, exhibit strong analogies to the time slices of threads in traditional operating systems. This insight presents a novel opportunity to tackle the problem at the system level. In this paper, we propose AMS, a robot action management system enhanced with OS-level primitives like exception, context switch and record-and-replay, that improves both execution efficiency and success rates of robotic tasks. AMS first introduces action exception, which facilitates the immediate interruption of robotic actions to prevent error propagation. Secondly, AMS proposes action context, which eliminates redundant computations for VLA-based models, thereby accelerating execution efficiency in robotic actions. Finally, AMS leverages action replay to facilitate repetitive or similar robotic tasks without the need for re-training efforts. We implement AMS in both an emulated environment and on a real robot platform. The evaluation results demonstrate that AMS significantly enhances the model's generalization ability and action efficiency, achieving task success rate improvements ranging from 7x to 24x and saving end-to-end execution time ranging from 29% to 74% compared to existing robotic system without AMS support.

cs.OS↗

arXiv2025

Topology-Aware Virtualization over Inter-Core Connected Neural Processing Units

With the rapid development of artificial intelligence (AI) applications, an emerging class of AI accelerators, termed Inter-core Connected Neural Processing Units (NPU), has been adopted in both cloud and edge computing environments, like Graphcore IPU, Tenstorrent, etc. Despite their innovative design, these NPUs often demand substantial hardware resources, leading to suboptimal resource utilization due to the imbalance of hardware requirements across various tasks. To address this issue, prior research has explored virtualization techniques for monolithic NPUs, but has neglected inter-core connected NPUs with the hardware topology. This paper introduces vNPU, the first comprehensive virtualization design for inter-core connected NPUs, integrating three novel techniques: (1) NPU route virtualization, which redirects instruction and data flow from virtual NPU cores to physical ones, creating a virtual topology; (2) NPU memory virtualization, designed to minimize translation stalls for SRAM-centric and NoC-equipped NPU cores, thereby maximizing the memory bandwidth; and (3) Best-effort topology mapping, which determines the optimal mapping from all candidate virtual topologies, balancing resource utilization with end-to-end performance. We have developed a prototype of vNPU on both an FPGA platform (Chipyard+FireSim) and a simulator (DCRA). Evaluation results indicate that, compared to other virtualization approaches such as unified virtual memory and MIG, vNPU achieves up to a 2x performance improvement across various ML models, with only 2% hardware cost.

cs.AR↗

arXiv2025

Get Experience from Practice: LLM Agents with Record & Replay

AI agents, empowered by Large Language Models (LLMs) and communication protocols such as MCP and A2A, have rapidly evolved from simple chatbots to autonomous entities capable of executing complex, multi-step tasks, demonstrating great potential. However, the LLMs' inherent uncertainty and heavy computational resource requirements pose four significant challenges to the development of safe and efficient agents: reliability, privacy, cost and performance. Existing approaches, like model alignment, workflow constraints and on-device model deployment, can partially alleviate some issues but often with limitations, failing to fundamentally resolve these challenges. This paper proposes a new paradigm called AgentRR (Agent Record & Replay), which introduces the classical record-and-replay mechanism into AI agent frameworks. The core idea is to: 1. Record an agent's interaction trace with its environment and internal decision process during task execution, 2. Summarize this trace into a structured "experience" encapsulating the workflow and constraints, and 3. Replay these experiences in subsequent similar tasks to guide the agent's behavior. We detail a multi-level experience abstraction method and a check function mechanism in AgentRR: the former balances experience specificity and generality, while the latter serves as a trust anchor to ensure completeness and safety during replay. In addition, we explore multiple application modes of AgentRR, including user-recorded task demonstration, large-small model collaboration and privacy-aware agent execution, and envision an experience repository for sharing and reusing knowledge to further reduce deployment cost.

cs.LG↗

arXiv2024

TensorTEE: Unifying Heterogeneous TEE Granularity for Efficient Secure Collaborative Tensor Computing

Heterogeneous collaborative computing with NPU and CPU has received widespread attention due to its substantial performance benefits. To ensure data confidentiality and integrity during computing, Trusted Execution Environments (TEE) is considered a promising solution because of its comparatively lower overhead. However, existing heterogeneous TEE designs are inefficient for collaborative computing due to fine and different memory granularities between CPU and NPU. 1) The cacheline granularity of CPU TEE intensifies memory pressure due to its extra memory access, and 2) the cacheline granularity MAC of NPU escalates the pressure on the limited memory storage. 3) Data transfer across heterogeneous enclaves relies on the transit of non-secure regions, resulting in cumbersome re-encryption and scheduling. To address these issues, we propose TensorTEE, a unified tensor-granularity heterogeneous TEE for efficient secure collaborative tensor computing. First, we virtually support tensor granularity in CPU TEE to eliminate the off-chip metadata access by detecting and maintaining tensor structures on-chip. Second, we propose tensor-granularity MAC management with predictive execution to avoid computational stalls while eliminating off-chip MAC storage and access. Moreover, based on the unified granularity, we enable direct data transfer without re-encryption and scheduling dilemmas. Our evaluation is built on enhanced Gem5 and a cycle-accurate NPU simulator. The results show that TensorTEE improves the performance of Large Language Model (LLM) training workloads by 4.0x compared to existing work and incurs only 2.1% overhead compared to non-secure training, offering a practical security assurance for LLM training.

cs.CR↗