arXiv subjects

Penghui Liu

Publications and source records attributed to Penghui Liu.

12 recordsLinked to original sources

arXiv2026

TrustMargin: Training-Free Arbitration between Parametric Memory and Retrieved Evidence in Large Language Models

Large language models answer knowledge-intensive questions using both parametric memory and retrieved evidence, but neither source is uniformly reliable. Retrieval can fill knowledge gaps, yet distracting passages may override correct closed-book answers. We study this post-generation conflict as answer-level source arbitration: given Direct and RAG answers from the same frozen model, decide which source to trust. We propose TRUSTMARGIN, a training-free, plug-and-play arbitration layer that scores the two existing candidates with the model's own likelihoods. It combines a parametric-prior margin, which tests whether memory accepts the retrieved answer, with an evidence-binding margin, which discounts passage-only salience and measures question-specific support. TRUSTMARGIN selects between Direct and RAG without fine-tuning, external judges, or additional generation. Across 2WIKIMQA and CWQA with three LLaMA scales, TRUSTMARGIN consistently improves over Direct generation and BM25-RAG, recovers part of the Direct/RAG oracle gap, and generalizes to multiple training-free RAG pipelines.

cs.CL↗

arXiv2026

ALAS: Adaptive Long-Horizon Action Synthesis via Async-pathway Stream Disentanglement

Long-Horizon (LH) tasks in Human-Scene Interaction (HSI) are complex multi-step tasks that require continuous planning, sequential decision-making, and extended execution across domains to achieve the final goal. However, existing methods heavily rely on skill chaining by concatenating pre-trained subtasks, with environment observations and self-state tightly coupled, lacking the ability to generalize to new combinations of environments and skills, failing to complete various LH tasks across domains. To solve this problem, this paper presents ALAS, a cross-domain learning framework for LH tasks via biologically inspired dual-stream disentanglement. Inspired by the brain's "where-what" dual pathway mechanism, ALAS comprises two core modules: i) an environment learning module for spatial understanding, which captures object functions, spatial relationships, and scene semantics, achieving cross-domain transfer through complete environment-self disentanglement; ii) a skill learning module for task execution, which processes self-state information including joint degrees of freedom and motor patterns, enabling cross-skill transfer through independent motor pattern encoding. We conducted extensive experiments on various LH tasks in HSI scenes. Compared with existing methods, ALAS can achieve an average subtasks success rate improvement of 23\% and average execution efficiency improvement of 29\%.

cs.RO↗

arXiv2026

Efficient ML-DSA Public Key Management Method with Identity for PKI and Its Application

With the rapid evolution of the Industrial Internet of Things (IIoT), the boundaries and scale of the Internet are continuously expanding. Consequently, the limitations of traditional certificate-based Public Key Infrastructure (PKI) have become increasingly evident, particularly in scenarios requiring large-scale certificate storage, verification, and frequent transmission. These challenges are expected to be further amplified by the widespread adoption of post-quantum cryptography. In this paper, we propose a novel identity-based public key management framework for PKI based on post-quantum cryptography, termed \textit{IPK-pq}. This approach implements an identity key generation protocol leveraging NIST ML-DSA and random matrix theory. Building on the concept of the Composite Public Key (CPK), \textit{IPK-pq} addresses the linear collusion problem inherent in CPK through an enhanced identity mapping mechanism. Furthermore, it simplifies the verification of the declared public key's authenticity, effectively reducing the complexity associated with certificate-based key management. We also provide a formal security proof for \textit{IPK-pq}, covering both individual private key components and the composite private key. To validate our approach, formally, we directly implement and evaluate \textit{IPK-pq} within a typical PKI application scenario: Resource PKI (RPKI). Comparative experimental results demonstrate that an RPKI system based on \textit{IPK-pq} yields significant improvements in efficiency and scalability. These results validate the feasibility and rationality of \textit{IPK-pq}, positioning it as a strong candidate for next-generation RPKI systems capable of securely managing large-scale routing information.

cs.CR↗

arXiv2026

MetaWorld-X: Hierarchical World Modeling via VLM-Orchestrated Experts for Humanoid Loco-Manipulation

Learning natural, stable, and compositionally generalizable whole-body control policies for humanoid robots performing simultaneous locomotion and manipulation (loco-manipulation) remains a fundamental challenge in robotics. Existing reinforcement learning approaches typically rely on a single monolithic policy to acquire multiple skills, which often leads to cross-skill gradient interference and motion pattern conflicts in high-degree-of-freedom systems. As a result, generated behaviors frequently exhibit unnatural movements, limited stability, and poor generalization to complex task compositions. To address these limitations, we propose MetaWorld-X, a hierarchical world model framework for humanoid control. Guided by a divide-and-conquer principle, our method decomposes complex control problems into a set of specialized expert policies (Specialized Expert Policies, SEP). Each expert is trained under human motion priors through imitation-constrained reinforcement learning, introducing biomechanically consistent inductive biases that ensure natural and physically plausible motion generation. Building upon this foundation, we further develop an Intelligent Routing Mechanism (IRM) supervised by a Vision-Language Model (VLM), enabling semantic-driven expert composition. The VLM-guided router dynamically integrates expert policies according to high-level task semantics, facilitating compositional generalization and adaptive execution in multi-stage loco-manipulation tasks.

cs.RO↗

arXiv2025

MultiMotion: Multi Subject Video Motion Transfer via Video Diffusion Transformer

Multi-object video motion transfer poses significant challenges for Diffusion Transformer (DiT) architectures due to inherent motion entanglement and lack of object-level control. We present MultiMotion, a novel unified framework that overcomes these limitations. Our core innovation is Maskaware Attention Motion Flow (AMF), which utilizes SAM2 masks to explicitly disentangle and control motion features for multiple objects within the DiT pipeline. Furthermore, we introduce RectPC, a high-order predictor-corrector solver for efficient and accurate sampling, particularly beneficial for multi-entity generation. To facilitate rigorous evaluation, we construct the first benchmark dataset specifically for DiT-based multi-object motion transfer. MultiMotion demonstrably achieves precise, semantically aligned, and temporally coherent motion transfer for multiple distinct objects, maintaining DiT's high quality and scalability. The code is in the supp.

cs.CV↗

arXiv2025

DETACH: Cross-domain Learning for Long-Horizon Tasks via Mixture of Disentangled Experts

Long-Horizon (LH) tasks in Human-Scene Interaction (HSI) are complex multi-step tasks that require continuous planning, sequential decision-making, and extended execution across domains to achieve the final goal. However, existing methods heavily rely on skill chaining by concatenating pre-trained subtasks, with environment observations and self-state tightly coupled, lacking the ability to generalize to new combinations of environments and skills, failing to complete various LH tasks across domains. To solve this problem, this paper presents DETACH, a cross-domain learning framework for LH tasks via biologically inspired dual-stream disentanglement. Inspired by the brain's "where-what" dual pathway mechanism, DETACH comprises two core modules: i) an environment learning module for spatial understanding, which captures object functions, spatial relationships, and scene semantics, achieving cross-domain transfer through complete environment-self disentanglement; ii) a skill learning module for task execution, which processes self-state information including joint degrees of freedom and motor patterns, enabling cross-skill transfer through independent motor pattern encoding. We conducted extensive experiments on various LH tasks in HSI scenes. Compared with existing methods, DETACH can achieve an average subtasks success rate improvement of 23% and average execution efficiency improvement of 29%.

cs.RO↗

arXiv2025

Multiple generation star formation in Cepheus Flare

We present an analysis of the young stellar moving group ASCC 127 using Gaia DR3 data, significantly expanding its membership to 3,971 stars -- double the number identified in previous studies. Using kinematic and distance criteria, ASCC 127 is divided into five subgroups (Groups 1-5) with ages spanning from 15 to 32 Myr. Groups 1-5 are spatially linked to the Cepheus Flare star-forming region, revealing potential evidence of four sequential star formation episodes at approximately 32 Myr, 20 Myr, 15 Myr, and 7 Myr. Through dust and gas mapping, we identify a spatial cavity extending several tens of parsecs, which may have resulted from feedback processes such as supernovae associated with earlier generations of stars in the region. This structure, along with the larger Loop III feature, indicates that feedback from massive stars likely influenced the interstellar medium (ISM). By integrating young stellar populations with ISM studies, we provide a detailed picture of the feedback-driven star formation history in the Cepheus Flare region.

astro-ph.SR↗

arXiv2025

Nearby open clusters with tidal features: golden sample selection and 3D structure

Open clusters offer unique opportunities to study stellar dynamics and evolution under the influence of their internal gravity, the Milky Way's gravitational field, and the interactions with encounters. Using the Gaia DR3 data for a catalog of open clusters within 500 parsecs that exhibit tidal features reported by the literature, we apply a novel method based on 3D principal component analysis to select a ``golden sample'' of nearby open clusters with minimal line-of-sight distortions. This approach ensures a systematic comparison of 3D and 2D structural parameters for tidally perturbed clusters. The selected golden sample includes Blanco 1, Melotte 20, Melotte 22, NGC 2632, NGC 7092, NGC 1662, Roslund 6 and Melotte 111. We analyze these clusters by fitting both 2D and 3D King Profiles to their stellar density distributions. Our results reveal systematic discrepancies: most of the golden sample clusters exhibit larger 3D tidal radii compared to their 2D counterparts, demonstrating that the 2D projection effects bias the measured cluster size. Furthermore, the 3D density profiles show stronger deviations from King profiles at the tidal radii ($\Delta \rho_{\rm 3D} > \Delta \rho_{\rm 2D}$), highlighting enhanced sensitivity to tidal disturbances. Additionally, we investigate the spatial distribution of cluster members relative to their bulk motion in the Galactic plane. We find that some clusters exhibit tidal features oriented perpendicular to their direction of motion, which can be attributed to the fact that the current surveys only detect the curved inner regions of the tidal features. In conclusion, this work offers a golden sample of nearby open clusters that are most reliable for 3D structure analysis and underscores the necessity of 3D analysis in characterizing OC morphological asymmetries, determining cluster size, and identifying tidal features.

astro-ph.GA↗

arXiv2025

Revisiting open clusters within 200 pc in the solar neighbourhood with Gaia DR3

In this study, we develop a membership identification method and apply it for 30 open clusters (OCs) within 200 pc of the Sun using astrometric data of Gaia DR3. By accounting for projection effects that distort apparent stellar motions, our approach converts astrometric data into accurate five-dimensional positions and velocities. This approach enables better identification of members in nearby open clusters. We then compare our refined membership lists with previous catalogs, revealing more members in most open clusters, but also the identification of elongated structures in Melotte 25 (Hyades), NGC 2632 (Praesepe), Melotte 111 (Coma Berenices), Platais 3, Melotte 22 (Pleiades), NGC 2451A, Platais 9, IC 2391, Platais 8, UPK 640, HSC 2986, which we studied in detail. An analysis of the ages of their members reveals the members within and outside of the tidal radius are distinctly coeval, further validating our methodology. This study suggests that for OCs in the solar neighborhood, correcting for the projection effect is very important for identification of OC members.

astro-ph.SR↗

arXiv2024

CRepair: CVAE-based Automatic Vulnerability Repair Technology

Software vulnerabilities are flaws in computer software systems that pose significant threats to the integrity, security, and reliability of modern software and its application data. These vulnerabilities can lead to substantial economic losses across various industries. Manual vulnerability repair is not only time-consuming but also prone to errors. To address the challenges of vulnerability repair, researchers have proposed various solutions, with learning-based automatic vulnerability repair techniques gaining widespread attention. However, existing methods often focus on learning more vulnerability data to improve repair outcomes, while neglecting the diverse characteristics of vulnerable code, and suffer from imprecise vulnerability localization.To address these shortcomings, this paper proposes CRepair, a CVAE-based automatic vulnerability repair technology aimed at fixing security vulnerabilities in system code. We first preprocess the vulnerability data using a prompt-based method to serve as input to the model. Then, we apply causal inference techniques to map the vulnerability feature data to probability distributions. By employing multi-sample feature fusion, we capture diverse vulnerability feature information. Finally, conditional control is used to guide the model in repairing the vulnerabilities.Experimental results demonstrate that the proposed method significantly outperforms other benchmark models, achieving a perfect repair rate of 52%. The effectiveness of the approach is validated from multiple perspectives, advancing AI-driven code vulnerability repair and showing promising applications.

cs.SE↗

arXiv2023

DECN: Evolution Inspired Deep Convolution Network for Black-box Optimization

Evolutionary algorithms (EAs) have emerged as a powerful framework for optimization, especially for black-box optimization. Existing evolutionary algorithms struggle to comprehend and effectively utilize task-specific information for adjusting their optimization strategies, leading to subpar performance on target tasks. Moreover, optimization strategies devised by experts tend to be highly biased. These challenges significantly impede the progress of the field of evolutionary computation. Therefore, this paper first introduces the concept of Automated EA: Automated EA exploits structure in the problem of interest to automatically generate update rules (optimization strategies) for generating and selecting potential solutions so that it can move a random population near the optimal solution. However, current EAs cannot achieve this goal due to the poor representation of the optimization strategy and the weak interaction between the optimization strategy and the target task. We design a deep evolutionary convolution network (DECN) to realize the move from hand-designed EAs to automated EAs without manual interventions. DECN has high adaptability to the target task and can obtain better solutions with less computational cost. DECN is also able to effectively utilize the low-fidelity information of the target task to form an efficient optimization strategy. The experiments on nine synthetics and two real-world cases show the advantages of learned optimization strategies over the state-of-the-art human-designed and meta-learning EA baselines. In addition, due to the tensorization of the operations, DECN is friendly to the acceleration provided by GPUs and runs 102 times faster than EA.

cs.NE↗

arXiv2023

Benchmarking Software Vulnerability Detection Techniques: A Survey

Software vulnerabilities can have serious consequences, which is why many techniques have been proposed to defend against them. Among these, vulnerability detection techniques are a major area of focus. However, there is a lack of a comprehensive approach for benchmarking these proposed techniques. In this paper, we present the first survey that comprehensively investigates and summarizes the current state of software vulnerability detection benchmarking. We review the current literature on benchmarking vulnerability detection, including benchmarking approaches in technique-proposing papers and empirical studies. We also separately discuss the benchmarking approaches for traditional and deep learning-based vulnerability detection techniques. Our survey analyzes the challenges of benchmarking software vulnerability detection techniques and the difficulties involved. We summarize the challenges of benchmarking software vulnerability detection techniques and describe possible solutions for addressing these challenges.

cs.SE↗