arXiv subjects

Xiaoliang Wang

Publications and source records attributed to Xiaoliang Wang.

At least 19 recordsLinked to original sources

arXiv2026

FenixWalk: Adaptive Sampling for Large-Scale Dynamic Graphs on Heterogeneous Architectures

Random walks are fundamental to graph analytics, yet executing them efficiently on large, evolving graphs remains challenging. Existing approaches must balance sampling efficiency, update overhead, and memory consumption. Sampling structures optimized for throughput often incur substantial memory overhead and costly maintenance, whereas lightweight alternatives can suffer from high sampling latency on high-degree vertices or skewed weight distributions. These challenges become more pronounced on heterogeneous CPU--GPU platforms, where host memory provides capacity but GPU memory limits the amount of graph data that can benefit from parallel acceleration. We present FenixWalk, a memory-efficient random-walk system that combines unified in-place sampling with selective GPU acceleration for large-scale dynamic graphs. At its core, FenixWalk employs an in-place Fenwick representation that supports both rejection sampling and inverse-transform sampling without duplicating edge-level weight storage. This representation enables expected O(1) weight lookup for rejection sampling, together with O(log d) prefix sampling and updates. To exploit heterogeneous hardware efficiently, FenixWalk maintains the complete graph and its sampling structures in host memory while selectively placing frequently reused vertex-local samplers on the GPU. By transferring compact walker states instead of repeatedly migrating graph partitions, the system reduces cross-device communication and makes effective use of limited GPU memory. Experiments on large real-world graphs and representative random-walk workloads demonstrate that FenixWalk achieves average speedups ranging from 2.06x to 400.57x over existing random-walk systems.

cs.DC↗

arXiv2026

MedEvo: Medical Reasoning Evolution through Knowledge Navigation

Medical multimodal reasoning models are typically trained on teacher-generated rationales, discarding the valid reasoning in students' own failed trajectories, while general-domain correction methods offer little guidance for errors in image reading, anatomical context, or differential diagnosis. We propose MedEvo, which uses medical knowledge, namely a tag taxonomy and error-type diagnosis, to navigate both trajectory repair and data selection. For each failed rollout, a gold-blind auditor locates the first consequential error and provides a minimal, answer-free hint; the same student retains the valid prefix and regenerates the erroneous suffix. After each round, MedEvo identifies capability gaps across medical tags on a held-out feedback set and prioritizes related questions for new rollouts and repairs. With supervised fine-tuning (SFT) alone, MedEvo raises Qwen3.5-9B mean accuracy across seven benchmarks from 65.57% to 69.12% (47.40% to 54.50% on the multimodal track of MedXpertQA (MedXpertQA-MM)), surpassing teacher-distillation SFT (66.93%). It also raises Qwen3.5-35B and Qwen3.5-397B mean accuracy from 67.91% to 73.04% and from 71.83% to 74.95%, respectively. Our analyses show that failed trajectories contain reasoning worth preserving, that repair success does not imply learning value, and that independent feedback is key to sustained improvement.

cs.AI↗

arXiv2026

ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution

Electrocardiogram (ECG) models need supervision that targets evolving weaknesses and remains faithful to waveform evidence. We introduce ECG-RSI, a recursive self-improvement framework combining verified waveform synthesis with feedback-guided intervention search. Signal remeasurement verifies waveform edits; construction errors guide generator repair, while learner errors motivate interventions compared on a separate selection set. Across seven benchmarks, three rounds raise Qwen3.5-35B-A3B's mean question-answering accuracy from 34.57% to 56.04% and mean classification F1 from 24.27 to 34.03. Across the retained trajectory, the preferred supervision shifts from direct reinforcement to confusion-focused contrasts. Continued feedback correction yields diminishing external gains. These results motivate separating error diagnosis from intervention selection and checking whether gains on the feedback set transfer to independent benchmarks.

cs.AI↗

arXiv2026

A Unified Arena for LLM Agents across Diverse Eurogame

Interactive environments offer a practical way to evaluate large language model (LLM) agents, and competitive games extend this evaluation to decisions shaped by competing goals and other agents' choices. Yet existing evaluations confined to individual games or execution frameworks leave unclear which competitive advantages persist across settings. We introduce DECE (Diverse Eurogame Competitive Environments), a unified arena spanning ten Eurogames with shared interfaces and evaluation protocols. These games combine lasting strategic commitments, competition over shared resources, and multiple routes to victory. Our primary study evaluates 12 model-harness configurations, spanning nine models and four execution harnesses, in 480 three-player matches across ten games. The main and supplementary studies comprise 1,032 completed matches in total. These comparisons reveal broad cross-game ranking agreement alongside game-specific strengths. Matched studies show that the same model can perform differently across execution harnesses, adding another source of variation to these profiles. These performance advantages do not consistently translate into better token or cost efficiency, making resource use an additional consideration when comparing configurations. Together, the findings support evaluating complete agent configurations by both their competitive performance across games and the resources needed to achieve it.

cs.AI↗

arXiv2026

On the Isoperimetric Inequality for the Third Robin Eigenvalue on Surfaces

We investigate the third Robin eigenvalue on simply connected domains of area $A$ in surfaces with Gaussian curvature bounded above by $K$. For $K>0$, we discover a transition at the critical parameter $α=-4π+KA$: two equal geodesic disks give a sharp upper bound when $α=-4π$ or $-4π+KA\leqα\leq4π$, but the two disk bound fails for $-4π<α<-4π+KA$. In the latter range, we establish a new upper bound involving a single geodesic disk with a modified Robin parameter and prove its sharpness in an explicit subrange. The sharp two-disk inequality persists for nonpositive curvature bounds, and Steklov--Helmholtz analogues are also obtained. Both the two-disk and one-disk bounds arise from a unified two-pole Green-function method, with the one-disk regime emerging from a degenerate area split that produces the $2π$ shift in the effective Robin parameter, while sharpness in the stated subrange is established by boundary oscillations that redistribute Robin boundary mass.

math.AP↗

arXiv2026

TrigReason: Trigger-Based Collaboration between Small and Large Reasoning Models

Large Reasoning Models (LRMs) achieve strong performance on complex tasks through extended chains of thought but suffer from high inference latency due to autoregressive reasoning. Recent work explores using Small Reasoning Models (SRMs) to accelerate LRM inference. In this paper, we systematically characterize the capability boundaries of SRMs and identify three common types of reasoning risks: (1) path divergence, where SRMs lack the strategic ability to construct an initial plan, causing reasoning to deviate from the most probable path; (2) cognitive overload, where SRMs fail to solve particularly difficult steps; and (3) recovery inability, where SRMs lack robust self-reflection and error correction mechanisms. To address these challenges, we propose TrigReason, a trigger-based collaborative reasoning framework that replaces continuous polling with selective intervention. TrigReason delegates most reasoning to the SRM and activates LRM intervention only when necessary-during initial strategic planning (strategic priming trigger), upon detecting extraordinary overconfidence (cognitive offload trigger), or when reasoning falls into unproductive loops (intervention request trigger). The evaluation results on AIME24, AIME25, and GPQA-D indicate that TrigReason matches the accuracy of full LRMs and SpecReason, while offloading 1.70x - 4.79x more reasoning steps to SRMs. Under edge-cloud conditions, TrigReason reduces latency by 43.9\% and API cost by 73.3\%. Our code is available at \href{https://github.com/QQQ-yi/TrigReason}{https://github.com/QQQ-yi/TrigReason}

cs.AI↗

arXiv2026

LifeBench: A Benchmark for Long-Horizon Multi-Source Memory

Long-term memory is fundamental for personalized agents capable of accumulating knowledge, reasoning over user experiences, and adapting across time. However, existing memory benchmarks primarily target declarative memory, specifically semantic and episodic types, where all information is explicitly presented in dialogues. In contrast, real-world actions are also governed by non-declarative memory, including habitual and procedural types, and need to be inferred from diverse digital traces. To bridge this gap, we introduce Lifebench, which features densely connected, long-horizon event simulation. It pushes AI agents beyond simple recall, requiring the integration of declarative and non-declarative memory reasoning across diverse and temporally extended contexts. Building such a benchmark presents two key challenges: ensuring data quality and scalability. We maintain data quality by employing real-world priors, including anonymized social surveys, map APIs, and holiday-integrated calendars, thus enforcing fidelity, diversity and behavioral rationality within the dataset. Towards scalability, we draw inspiration from cognitive science and structure events according to their partonomic hierarchy; enabling efficient parallel generation while maintaining global coherence. Performance results show that top-tier, state-of-the-art memory systems reach just 55.2\% accuracy, highlighting the inherent difficulty of long-horizon retrieval and multi-source integration within our proposed benchmark. The dataset and data synthesis code are available at https://github.com/1754955896/LifeBench.

cs.AI↗

arXiv2026

SmallKV: Small Model Assisted Compensation of KV Cache Compression for Efficient LLM Inference

KV cache eviction has emerged as an effective solution to alleviate resource constraints faced by LLMs in long-context scenarios. However, existing token-level eviction methods often overlook two critical aspects: (1) their irreversible eviction strategy fails to adapt to dynamic attention patterns during decoding (the saliency shift problem), and (2) they treat both marginally important tokens and truly unimportant tokens equally, despite the collective significance of marginal tokens to model performance (the marginal information over-compression problem). To address these issues, we design two compensation mechanisms based on the high similarity of attention matrices between LLMs of different scales. We propose SmallKV, a small model assisted compensation method for KV cache compression. SmallKV can maintain attention matching between different-scale LLMs to: 1) assist the larger model in perceiving globally important information of attention; and 2) use the smaller model's attention scores to approximate those of marginal tokens in the larger model. Extensive experiments on benchmarks including GSM8K, BBH, MT-Bench, and LongBench demonstrate the effectiveness of SmallKV. Moreover, efficiency evaluations show that SmallKV achieves 1.75 - 2.56 times higher throughput than baseline methods, highlighting its potential for efficient and performant LLM inference in resource constrained environments.

cs.LG↗

arXiv2025

LAVa: Layer-wise KV Cache Eviction with Dynamic Budget Allocation

KV Cache is commonly used to accelerate LLM inference with long contexts, yet its high memory demand drives the need for cache compression. Existing compression methods, however, are largely heuristic and lack dynamic budget allocation. To address this limitation, we introduce a unified framework for cache compression by minimizing information loss in Transformer residual streams. Building on it, we analyze the layer attention output loss and derive a new metric to compare cache entries across heads, enabling layer-wise compression with dynamic head budgets. Additionally, by contrasting cross-layer information, we also achieve dynamic layer budgets. LAVa is the first unified strategy for cache eviction and dynamic budget allocation that, unlike prior methods, does not rely on training or the combination of multiple strategies. Experiments with benchmarks (LongBench, Needle-In-A-Haystack, Ruler, and InfiniteBench) demonstrate its superiority. Moreover, our experiments reveal a new insight: dynamic layer budgets are crucial for generation tasks (e.g., code completion), while dynamic head budgets play a key role in extraction tasks (e.g., extractive QA). As a fully dynamic compression method, LAVa consistently maintains top performance across task types. Our code is available at https://github.com/MGDDestiny/Lava.

cs.LG↗

arXiv2025

Consultant Decoding: Yet Another Synergistic Mechanism

The synergistic mechanism based on Speculative Decoding (SD) has garnered considerable attention as a simple yet effective approach for accelerating the inference of large language models (LLMs). Nonetheless, the high rejection rates require repeated LLMs calls to validate draft tokens, undermining the overall efficiency gain of SD. In this work, we revisit existing verification mechanisms and propose a novel synergetic mechanism Consultant Decoding (CD). Unlike SD, which relies on a metric derived from importance sampling for verification, CD verifies candidate drafts using token-level likelihoods computed solely by the LLM. CD achieves up to a 2.5-fold increase in inference speed compared to the target model, while maintaining comparable generation quality (around 100% of the target model's performance). Interestingly, this is achieved by combining models whose parameter sizes differ by two orders of magnitude. In addition, CD reduces the call frequency of the large target model to below 10%, particularly in more demanding tasks. CD's performance was even found to surpass that of the large target model, which theoretically represents the upper bound for speculative decoding.

cs.CL↗

arXiv2025

daDPO: Distribution-Aware DPO for Distilling Conversational Abilities

Large language models (LLMs) have demonstrated exceptional performance across various applications, but their conversational abilities decline sharply as model size decreases, presenting a barrier to their deployment in resource-constrained environments. Knowledge distillation with Direct Preference Optimization (dDPO) has emerged as a promising approach to enhancing the conversational abilities of smaller models using a larger teacher model. However, current methods primarily focus on 'black-box' KD, which only uses the teacher's responses, overlooking the output distribution offered by the teacher. This paper addresses this gap by introducing daDPO (Distribution-Aware DPO), a unified method for preference optimization and distribution-based distillation. We provide rigorous theoretical analysis and empirical validation, showing that daDPO outperforms existing methods in restoring performance for pruned models and enhancing smaller LLM models. Notably, in in-domain evaluation, our method enables a 20% pruned Vicuna1.5-7B to achieve near-teacher performance (-7.3% preference rate compared to that of dDPO's -31%), and allows Qwen2.5-1.5B to occasionally outperform its 7B teacher model (14.0% win rate).

cs.LG↗

arXiv2025

Semiparametric Conditional Factor Models in Asset Pricing

We introduce a simple and tractable methodology for estimating semiparametric conditional latent factor models. Our approach disentangles the roles of characteristics in capturing factor betas of asset returns from ``alpha.'' We construct factors by extracting principal components from Fama-MacBeth managed portfolios. Applying this methodology to the cross-section of U.S. individual stock returns, we find compelling evidence of substantial nonzero pricing errors, even though our factors demonstrate superior performance in standard asset pricing tests. Unexplained ``arbitrage'' portfolios earn high Sharpe ratios, which decline over time. Combining factors with these orthogonal portfolios produces out-of-sample Sharpe ratios exceeding 4.

econ.EM↗

arXiv2024

SynthDoc: Bilingual Documents Synthesis for Visual Document Understanding

This paper introduces SynthDoc, a novel synthetic document generation pipeline designed to enhance Visual Document Understanding (VDU) by generating high-quality, diverse datasets that include text, images, tables, and charts. Addressing the challenges of data acquisition and the limitations of existing datasets, SynthDoc leverages publicly available corpora and advanced rendering tools to create a comprehensive and versatile dataset. Our experiments, conducted using the Donut model, demonstrate that models trained with SynthDoc's data achieve superior performance in pre-training read tasks and maintain robustness in downstream tasks, despite language inconsistencies. The release of a benchmark dataset comprising 5,000 image-text pairs not only showcases the pipeline's capabilities but also provides a valuable resource for the VDU community to advance research and development in document image recognition. This work significantly contributes to the field by offering a scalable solution to data scarcity and by validating the efficacy of end-to-end models in parsing complex, real-world documents.

cs.CV↗

arXiv2024

HardTaint: Production-Run Dynamic Taint Analysis via Selective Hardware Tracing

Dynamic taint analysis (DTA), as a fundamental analysis technique, is widely used in security, privacy, and diagnosis, etc. As DTA demands to collect and analyze massive taint data online, it suffers extremely high runtime overhead. Over the past decades, numerous attempts have been made to lower the overhead of DTA. Unfortunately, the reductions they achieved are marginal, causing DTA only applicable to the debugging/testing scenarios. In this paper, we propose and implement HardTaint, a system that can realize production-run dynamic taint tracking. HardTaint adopts a hybrid and systematic design which combines static analysis, selective hardware tracing and parallel graph processing techniques. The comprehensive evaluations demonstrate that HardTaint introduces only around 9% runtime overhead which is an order of magnitude lower than the state-of-the-arts, while without sacrificing any taint detection capability.

cs.CR↗

arXiv2023

Secure Inter-domain Routing and Forwarding via Verifiable Forwarding Commitments

The Internet inter-domain routing system is vulnerable. On the control plane, the de facto Border Gateway Protocol (BGP) does not have built-in mechanisms to authenticate routing announcements, so an adversary can announce virtually arbitrary paths to hijack network traffic; on the data plane, it is difficult to ensure that actual forwarding path complies with the control plane decisions. The community has proposed significant research to secure the routing system. Yet, existing secure BGP protocols (e.g., BGPsec) are not incrementally deployable, and existing path authorization protocols are not compatible with the current Internet routing infrastructure. In this paper, we propose FC-BGP, the first secure Internet inter-domain routing system that can simultaneously authenticate BGP announcements and validate data plane forwarding in an efficient and incrementally-deployable manner. FC-BGP is built upon a novel primitive, name Forwarding Commitment, to certify an AS's routing intent on its directly connected hops. We analyze the security benefits of FC-BGP in the Internet at different deployment rates. Further, we implement a prototype of FC-BGP and extensively evaluate it over a large-scale overlay network with 100 virtual machines deployed globally. The results demonstrate that FC-BGP saves roughly 55% of the overhead required to validate BGP announcements compared with BGPsec, and meanwhile FC-BGP introduces a small overhead for building a globally-consistent view on the desirable forwarding paths.

cs.NI↗

arXiv2023

From RDMA to RDCA: Toward High-Speed Last Mile of Data Center Networks Using Remote Direct Cache Access

In this paper, we conduct systematic measurement studies to show that the high memory bandwidth consumption of modern distributed applications can lead to a significant drop of network throughput and a large increase of tail latency in high-speed RDMA networks.We identify its root cause as the high contention of memory bandwidth between application processes and network processes. This contention leads to frequent packet drops at the NIC of receiving hosts, which triggers the congestion control mechanism of the network and eventually results in network performance degradation. To tackle this problem, we make a key observation that given the distributed storage service, the vast majority of data it receives from the network will be eventually written to high-speed storage media (e.g., SSD) by CPU. As such, we propose to bypass host memory when processing received data to completely circumvent this performance bottleneck. In particular, we design Lamda, a novel receiver cache processing system that consumes a small amount of CPU cache to process received data from the network at line rate. We implement a prototype of Lamda and evaluate its performance extensively in a Clos-based testbed. Results show that for distributed storage applications, Lamda improves network throughput by 4.7% with zero memory bandwidth consumption on storage nodes, and improves network throughput by up 17% and 45% for large block size and small size under the memory bandwidth pressure, respectively. Lamda can also be applied to latency-sensitive HPC applications, which reduces their communication latency by 35.1%.

cs.NI↗

arXiv2022

Chemical principles of instability and self-organization in reacting and diffusive systems

How patterns and structures undergo symmetry breaking and self-organize within biological systems from initially homogeneous states is a key issue for biological development. The activator-inhibitor (AI) mechanism, derived from reaction-diffusion (RD) models, has been widely believed to be the elementary mechanism for biological pattern formation. This mechanism generally requires activators to be self-enhanced and diffuse more slowly than inhibitors. Here, we identify the instability sources of biological systems and derive the self-organization conditions through solving eigenvalues (dispersion relation) of the generalized RD model for two chemicals. We show that both the single AI mechanisms with long-range inhibition and activation are enough to self-organize into fully-expressed domains without the involvement of the inhibitor-inhibitor (II) mechanism, through singly enhancing the difference in self-proliferation rates of activators and inhibitors or weakening the coupling degree between them. When cross diffusion involves, both the self-enhancement and the difference in diffusion coefficients of chemicals are no longer necessary for self-organization, and the patterning mechanism can be extended to semi-inhibitor and II mechanisms. However, we show that the single activator-activator (AA) mechanism is generally unable to self-organize, even if biological domain growth is additionally involved. Moreover, adding an II system after an AI one can produce discrete and bi-stable patterns. We also observe that a higher dimensional space can solely alter the patterning principles derived from a lower dimensional space, which may be due to the instability driven by the higher degree of spatial freedom. Such results provide new insights into biological pattern formation.

physics.bio-ph↗

arXiv2022

ALT: Boosting Deep Learning Performance by Breaking the Wall between Graph and Operator Level Optimizations

Deep learning models rely on highly optimized tensor libraries for efficient inference on heterogeneous hardware. Current deep compilers typically predetermine layouts of tensors and then optimize loops of operators. However, such unidirectional and one-off workflow strictly separates graph-level optimization and operator-level optimization into different system layers, missing opportunities for unified tuning. This paper proposes ALT, a compiler that performs joint graph- and operator-level optimizations for deep models. ALT provides a generic transformation module to manipulate layouts and loops with easy-to-use primitive functions. ALT further integrates an auto-tuning module that jointly optimizes graph-level data layouts and operator-level loops while guaranteeing efficiency. Experimental results show that ALT significantly outperforms state-of-the-art compilers (e.g., Ansor) in terms of both single operator performance (e.g., 1.5x speedup on average) and end-to-end inference performance (e.g., 1.4x speedup on average).

cs.LG↗