arXiv subjects

Xianwei Zhang

Publications and source records attributed to Xianwei Zhang.

13 recordsLinked to original sources

arXiv2026

Benchmarks are Not Enough: RAMP for Runtime Assessing of Agentic Models in Production Systems

LLM agents are rapidly evolving from coding assistants into autonomous software engineering systems. However, existing evaluation methodologies remain largely centered on static, isolated, and short-horizon benchmarks that fail to capture the dynamic complexity of real-world production workflows. As a result, benchmark performance may poorly reflect practical capability under realistic runtime environments involving long execution chains, tool interactions, dependency management, and iterative feedback loops. We thus present RAMP, a production-grounded infrastructure for assessing long-horizon software engineering agents. Built upon the YatCC integrated platform, RAMP provides a unified runtime assessment architecture through standardized orchestration and execution interfaces. RAMP introduces realistic compiler-construction workloads with serial dependencies and complex toolchain interactions, together with a staged recovery mechanism for analyzing execution behavior under partial workflow failure. The framework further incorporates utility-oriented multi-dimensional metrics that jointly evaluate outcome quality and process efficiency. We conduct runtime assessments across 15 mainstream models and observe substantial capability degradation that remains largely invisible to conventional isolated benchmarks. Task completion rates progressively collapse across serial workflows, dropping from 100% in the initial stage to only 20% in the final stage, while none of the evaluated models successfully completes the entire pipeline. Runtime analysis reveals systematic failure propagation and significant resource inefficiencies, with computational costs differing by up to three orders of magnitude among comparable models. These findings suggest RAMP advances agentic model evaluation toward continuous, runtime-observable, and production-grounded assessment.

cs.SE↗

arXiv2026

StepAudio 2.5 Technical Report

Unified audio-language modeling has emerged as a prominent trend in modern speech systems, promising to bring the reasoning capabilities of large language models to auditory tasks. However, existing unified foundations often struggle to match the depth of specialized systems across automatic speech recognition (ASR), text-to-speech synthesis (TTS), and realtime spoken interaction. Bridging this gap remains an open challenge. This report presents StepAudio 2.5, a unified audio-language foundation model that matches or exceeds specialized systems across all three capabilities. Rather than treating these tasks as architecturally distinct, we operate on the premise that once text and audio share a multimodal representational space, task specialization becomes a matter of operational regimes: data construction, optimization targets, and decoding constraints. Guided by this insight, we advance the post-training paradigm from standard supervised learning to task-tailored Reinforcement Learning from Human Feedback (RLHF), using it as the primary mechanism to define complex optimization targets. We leverage this RLHF-centric alignment, alongside specialized decoding, to shape a shared backbone into three distinct operational modes. Concretely, the ASR branch advances transcription efficiency via verifiable multi-token decoding; the TTS branch achieves controllable, expressive synthesis through preference-based RLHF and context-rich supervision; and the Realtime branch realizes low-latency, persona-consistent dialogue via generative reward modeling within an RLHF framework. On standard benchmarks, StepAudio 2.5 achieves state-of-the-art results across ASR, TTS, and Realtime, demonstrating that a singular audio-language foundation can successfully internalize the distinct deployment objectives of speech understanding, generation, and live interaction.

eess.AS↗

arXiv2026

SkCC: Portable and Secure Skill Compilation for Cross-Framework LLM Agents

LLM agents increasingly rely on reusable skills (e.g., SKILL markdown files) to execute complex tasks, yet these artifacts lack portability: agent frameworks are highly sensitive to prompt formatting, leading to a large performance variation for the same skill. Nevertheless, most skills are authored once as format-agnostic Markdown, necessitating costly per-framework rewrites and also leaving security largely unaddressed, with widespread vulnerabilities in practice. To address this, we present SkCC, a compiler for LLM agents that introduces classical compilation design into agent skill development. SkCC centers on SkIR, a strongly-typed intermediate representation that decouples skill semantics from framework-specific formatting, thus enabling portable deployment across agent frameworks. Atop of this IR, a static Optimizer enforces security constraints, blocking vulnerabilities before deployment. Implemented as a four-phase pipeline, SkCC effectively reduces adaptation complexity from $O(m \times n)$ to $O(m + n)$ across $m$ skills and $n$ frameworks. Experiments on SkillsBench demonstrate that SkCC delivers consistent and substantial gains over original counterparts, with pass rate increases from 21.1% to 33.3% on Claude Code and from 35.1% to 48.7% on Kimi CLI. Further, the design achieves sub-10ms compilation latency, 94.8% proactive security trigger rate, and 10-46% runtime token savings across frameworks.

cs.CR↗

arXiv2025

Partial Cross-Compilation and Mixed Execution for Accelerating Dynamic Binary Translation

With the growing diversity of instruction set architectures (ISAs), cross-ISA program execution has become common. Dynamic binary translation (DBT) is the main solution but suffers from poor performance. Cross-compilation avoids emulation costs but is constrained by an "all-or-nothing" model-programs are either fully cross-compiled or entirely emulated. Complete cross-compilation is often unfeasible due to ISA-specific code or missing dependencies, leaving programs with high emulation overhead. We propose a hybrid execution system that combines compilation and emulation, featuring a selective function offloading mechanism. This mechanism establishes cross-environment calling channels, offloading eligible functions to the host for native execution to reduce DBT overhead. Key optimizations address offloading costs, enabling efficient hybrid operation. Built on LLVM and QEMU, the system works automatically for both applications and libraries. Evaluations show it achieves up to 13x speedups over existing DBT, with strong practical value.

cs.AR↗

arXiv2025

Opt4GPTQ: Co-Optimizing Memory and Computation for 4-bit GPTQ Quantized LLM Inference on Heterogeneous Platforms

The increasing adoption of large language models (LLMs) on heterogeneous computing platforms poses significant challenges to achieving high inference efficiency. To address these efficiency bottlenecks across diverse platforms, this paper proposes Opt4GPTQ, a practical optimization method designed for 4-bit GPTQ quantized LLMs inference on heterogeneous AI accelerators. Built upon the vLLM serving system, Opt4GPTQ integrates three platform-level optimization strategies: Shared Memory Buffering Optimization (SMB-Opt), which caches frequently accessed data in shared memory and employs single-threaded writes; Vectorized Memory Loading Optimization (VML-Opt), which utilizes vectorized memory operations for efficient data loading; and Inline Assembly Optimization (ILA-Opt), which directly leverages hardwarenative vector half-precision addition and fused multiply-accumulate instructions. Experimental results show that Opt4GPTQ effectively improves performance across various models while maintaining original model accuracy, achieving throughput gains of up to 84.42%. This work highlights the critical role of platformlevel engineering in enabling efficient LLMs inference on emerging architectures and provides valuable methodologies for future heterogeneous platform adaptation.

cs.DC↗

arXiv2025

RServe: Overlapping Encoding and Prefill for Efficient LMM Inference

Large multimodal models (LMMs) typically employ an encoding module to transform multimodal data inputs into embeddings, which are then fed to language models for further processing. However, efficiently serving LMMs remains highly challenging due to the inherent complexity of their inference pipelines. Traditional serving engines co-locate the encoding module and the language model, leading to significant resource interference and tight data dependency. Recent studies have alleviated this issue by disaggregating the encoding module from the model, following a design style of prefill-decode disaggregation. Nevertheless, these approaches fail to fully exploit parallelism both within individual requests (intra-request) and across multiple requests (inter-request). To overcome the limitation, we propose REDServe, an LMM inference system that efficiently orchestrates intra- and inter-request pipelines. REDServe is designed to reduce low latency and maximize parallelism at both intra- and inter-request granularities. Built on the disaggregated architecture of the encoding module and language model, REDServe adopts a fine-grained scheduling method that overlaps multimodal encoding with the forward computation of the language model within a single request. For inter-request pipeline, REDServe leverages schedulable tokens and token budgets to balance computational loads across micro-batches. Combined with chunked prefill, this enables a novel scheduling strategy that coordinates the execution of intra- and inter-request pipelines. Experimental evaluations on representative LMMs show that REDServe achieves substantial latency reduction of up to 66% while improving throughput by up to 109%, significantly outperforming existing serving approaches.

cs.DC↗

arXiv2025

Boosting LLM Serving through Spatial-Temporal GPU Resource Sharing

Modern LLM serving systems confront inefficient GPU utilization due to the fundamental mismatch between compute-intensive prefill and memory-bound decode phases. While current practices attempt to address this by organizing these phases into hybrid batches, such solutions create an inefficient tradeoff that sacrifices either throughput or latency, leaving substantial GPU resources underutilized. We identify two key root causes: 1) the prefill phase suffers from suboptimal compute utilization due to wave quantization and attention bottlenecks. 2) hybrid batches disproportionately prioritize latency over throughput, resulting in wasted compute and memory bandwidth. To mitigate the issues, we present Bullet, a novel spatial-temporal orchestration system that eliminates these inefficiencies through precise phase coordination. Bullet enables concurrent execution of prefill and decode phases, while dynamically provisioning GPU resources using real-time performance modeling. By integrating SLO-aware scheduling and adaptive resource allocation, Bullet maximizes utilization without compromising latency targets. Experimental evaluations on real-world workloads demonstrate that Bullet delivers 1.26x average throughput gains (up to 1.55x) over state-of-the-arts, while consistently meeting latency constraints.

cs.DC↗

arXiv2025

Accelerating Edge Inference for Distributed MoE Models with Latency-Optimized Expert Placement

The emergence of Mixture-of-Experts (MoE) has transformed the scaling of large language models by enabling vast model capacity through sparse activation. Yet, converting these performance gains into practical edge deployment remains difficult, as the massive memory footprint and communication demands often overwhelm resource-limited environments. While centralized cloud-based solutions are available, they are frequently plagued by prohibitive infrastructure costs, latency issues, and privacy concerns. Moreover, existing edge-oriented optimizations largely overlook the complexities of heterogeneous hardware, focusing instead on isolated or uniform device setups. In response, this paper proposes Prism, an inference framework engineered for collaborative MoE serving across diverse GPU-equipped edge servers. By leveraging the intrinsic sparsity and input locality of MoE workloads, Prism minimizes inter-server communication and optimizes expert placement within diverse resource constraints. The framework integrates an activation-aware placement strategy that balances local request coverage with memory utilization, supplemented by a runtime migration mechanism to adapt expert distribution to dynamic workload changes. Experiments on contemporary MoE models and datasets demonstrate that Prism reduces inference latency by up to 30.6% and significantly lowers communication costs compared to state-of-the-art baselines, confirming the effectiveness of cooperative edge-based MoE serving.

cs.DC↗

arXiv2025

VecTrans: Enhancing Compiler Auto-Vectorization through LLM-Assisted Code Transformations

Auto-vectorization is a fundamental optimization for modern compilers to exploit SIMD parallelism. However, state-of-the-art approaches still struggle to handle intricate code patterns, often requiring manual hints or domain-specific expertise. Large language models (LLMs), with their ability to capture intricate patterns, provide a promising solution, yet their effective application in compiler optimizations remains an open challenge due to issues such as hallucinations and a lack of domain-specific reasoning. In this paper, we present VecTrans, a novel framework that leverages LLMs to enhance compiler-based code vectorization. VecTrans first employs compiler analysis to identify potentially vectorizable code regions. It then utilizes an LLM to refactor these regions into patterns that are more amenable to the compilers auto-vectorization. To ensure semantic correctness, VecTrans further integrates a hybrid validation mechanism at the intermediate representation (IR) level. With the above efforts, VecTrans combines the adaptability of LLMs with the precision of compiler vectorization, thereby effectively opening up the vectorization opportunities. experimental results show that among all TSVC functions unvectorizable by GCC, ICC, Clang, and BiSheng Compiler, VecTrans achieves an geomean speedup of 1.77x and successfully vectorizes 24 of 51 test cases. This marks a significant advancement over state-of-the-art approaches while maintaining a cost efficiency of $0.012 per function optimization for LLM API usage.

cs.SE↗

arXiv2025

EFIM: Efficient Serving of LLMs for Infilling Tasks with Improved KV Cache Reuse

Large language models (LLMs) are often used for infilling tasks, which involve predicting or generating missing information in a given text. These tasks typically require multiple interactions with similar context. To reduce the computation of repeated historical tokens, cross-request key-value (KV) cache reuse, a technique that stores and reuses intermediate computations, has become a crucial method in multi-round interactive services. However, in infilling tasks, the KV cache reuse is often hindered by the structure of the prompt format, which typically consists of a prefix and suffix relative to the insertion point. Specifically, the KV cache of the prefix or suffix part is frequently invalidated as the other part (suffix or prefix) is incrementally generated. To address the issue, we propose EFIM, a transformed prompt format of FIM to unleash the performance potential of KV cache reuse. Although the transformed prompt can solve the inefficiency, it exposes subtoken generation problems in current LLMs, where they have difficulty generating partial words accurately. Therefore, we introduce a fragment tokenization training method which splits text into multiple fragments before tokenization during data processing. Experiments on two representative LLMs show that LLM serving with EFIM can lower the latency by 52% and improve the throughput by 98% while maintaining the original infilling capability. EFIM's source code is publicly available at https://github.com/gty111/EFIM.

cs.CL↗

arXiv2025

gLLM: Global Balanced Pipeline Parallelism System for Distributed LLM Serving with Token Throttling

Pipeline parallelism has emerged as a predominant approach for deploying large language models (LLMs) across distributed nodes, owing to its lower communication overhead compared to tensor parallelism. While demonstrating high throughput in request serving, pipeline parallelism often suffers from performance limitations caused by pipeline bubbles, which are primarily resulted from imbalanced computation delays across batches. Existing methods like Sarathi-Serve attempt to address this through hybrid scheduling of chunked prefill and decode tokens using a fixed token budget. However, such methods may experience significant fluctuations due to either insufficient prefill tokens or uneven distribution of decode tokens, ultimately leading to computational imbalance. To overcome these inefficiencies, we present gLLM, a globally balanced pipeline parallelism system incorporating Token Throttling to effectively mitigate the pipeline bubbles. Our Token Throttling mechanism is a fine-grained scheduling policy that independently regulates the quantities of prefill and decode tokens, thus enabling balanced computation by leveraging global information from the inference system. Specifically, for decode tokens, gLLM maintains near-consistent token count across processing batches. For prefill tokens, it dynamically adjusts batch sizes based on both total pending tokens and the memory utilization rates of key-value cache (KV cache). Furthermore, gLLM runtime adopts an asynchronous execution and message passing architecture specifically optimized for pipeline parallelism characteristics. Experimental evaluations with representative LLMs show that gLLM achieves significant performance improvements, delivering 11% to 398% higher maximum throughput compared to state-of-the-art pipeline or tensor parallelism systems, while simultaneously maintaining lower latency.

cs.DC↗

arXiv2019

THUEE system description for NIST 2019 SRE CTS Challenge

This paper describes the systems submitted by the department of electronic engineering, institute of microelectronics of Tsinghua university and TsingMicro Co. Ltd. (THUEE) to the NIST 2019 speaker recognition evaluation CTS challenge. Six subsystems, including etdnn/ams, ftdnn/as, eftdnn/ams, resnet, multitask and c-vector are developed in this evaluation.

cs.SD↗

arXiv2019

Optimizing GPU Cache Policies for MI Workloads

In recent years, machine intelligence (MI) applications have emerged as a major driver for the computing industry. Optimizing these workloads is important but complicated. As memory demands grow and data movement overheads increasingly limit performance, determining the best GPU caching policy to use for a diverse range of MI workloads represents one important challenge. To study this, we evaluate 17 MI applications and characterize their behaviors using a range of GPU caching strategies. In our evaluations, we find that the choice of caching policy in GPU caches involves multiple performance trade-offs and interactions, and there is no one-size-fits-all GPU caching policy for MI workloads. Based on detailed simulation results, we motivate and evaluate a set of cache optimizations that consistently match the performance of the best static GPU caching policies.

cs.AR↗