arXiv subjects

Kun Hu

Publications and source records attributed to Kun Hu.

At least 37 records · Page 2Linked to original sources

arXiv2026

CABTO: Context-Aware Behavior Tree Grounding for Robot Manipulation

Behavior Trees (BTs) offer a powerful paradigm for designing modular and reactive robot controllers. BT planning, an emerging field, provides theoretical guarantees for the automated generation of reliable BTs. However, BT planning typically assumes that a well-designed BT system is already grounded -- comprising high-level action models and low-level control policies -- which often requires extensive expert knowledge and manual effort. In this paper, we formalize the BT Grounding problem: the automated construction of a complete and consistent BT system. We analyze its complexity and introduce CABTO (Context-Aware Behavior Tree grOunding), the first framework to efficiently solve this challenge. CABTO leverages pre-trained Large Models (LMs) to heuristically search the space of action models and control policies, guided by contextual feedback from BT planners and environmental observations. Experiments spanning seven task sets across three distinct robotic manipulation scenarios demonstrate CABTO's effectiveness and efficiency in generating complete and consistent behavior tree systems.

cs.RO↗

arXiv2026

OrthoEraser: Coupled-Neuron Orthogonal Projection for Concept Erasure

Text-to-image (T2I) models face significant safety risks from adversarial induction, yet current concept erasure methods often cause collateral damage to benign attributes when suppressing selected neurons entirely. This occurs because sensitive and benign semantics exhibit non-orthogonal superposition, sharing activation subspaces where their respective vectors are inherently entangled. To address this issue, we propose OrthoEraser, which leverages sparse autoencoders (SAE) to achieve high-resolution feature disentanglement and subsequently redefines erasure as an analytical orthogonalization projection that preserves the benign manifold's invariance. OrthoEraser first employs SAE to decompose dense activations and segregate sensitive neurons. It then uses coupled neuron detection to identify non-sensitive features vulnerable to intervention. The key novelty lies in an analytical gradient orthogonalization strategy that projects erasure vectors onto the null space of the coupled neurons. This orthogonally decouples the sensitive concepts from the identified critical benign subspace, effectively preserving non-sensitive semantics. Experimental results on safety demonstrate that OrthoEraser achieves high erasure precision, effectively removing harmful content while preserving the integrity of the generative manifold, and significantly outperforming SOTA baselines. This paper contains results of unsafe models.

cs.CV↗

arXiv2026

Enhancing Requirements Traceability Link Recovery: A Novel Approach with T-SimCSE

Requirements traceability plays an important role in ensuring software quality and responding to changes in requirements. Requirements trace links (such as the links between requirements and other software artifacts) underpin the modeling and implementation of requirements traceability. With the rapid development of artificial intelligence, more and more pre-trained language models (PLMs) techniques are applied to the automatic recovery of requirements trace links. However, the requirements traceability links recovered by these approaches are not accurate enough, and many approaches require a large labeled dataset for training. Currently, there are very few labeled datasets available. To address these limitations, this paper proposes a novel requirements traceability link recovery approach called T-SimCSE, which is based on a PLM -- SimCSE. SimCSE has the advantages of not requiring labeled data, having broad applicability, and performing well. T-SimCSE firstly uses the SimCSE model to calculate the similarity between requirements and target artifacts, and employs a new metric (i.e. specificity) to reorder those target artifacts. Finally, the trace links are created between the requirement and the top-K target artifacts. We have evaluated T-SimCSE on ten public datasets by comparing them with other approaches. The results show that T-SimCSE achieves superior performance in terms of recall and Mean Average Precision (MAP).

cs.SE↗

arXiv2026

Charting Empirical Laws for LLM Fine-Tuning in Scientific Multi-Discipline Learning

While large language models (LLMs) have achieved strong performance through fine-tuning within individual scientific domains, their learning dynamics in multi-disciplinary contexts remains poorly understood, despite the promise of improved generalization and broader applicability through cross-domain knowledge synergy. In this work, we present the first systematic study of multi-disciplinary LLM fine-tuning, constructing a five-discipline corpus and analyzing learning patterns of full fine-tuning, LoRA, LoRA-MoE, and LoRA compositions. Particularly, our study shows that multi-disciplinary learning is substantially more variable than single-discipline training and distills four consistent empirical laws: (1) Balance-then-Diversity: low-resource disciplines degrade performance unless mitigated via diversity-aware upsampling; (2) Merge-then-Align: restoring instruction-following ability is critical for cross-discipline synergy; (3) Optimize-then-Scale: parameter scaling offers limited gains without prior design optimization; and (4) Share-then-Specialize: asymmetric LoRA-MoE yields robust gains with minimal trainable parameters via shared low-rank projection. Together, these laws form a practical recipe for principled multi-discipline fine-tuning and provide actionable guidance for developing generalizable scientific LLMs.

cs.LG↗

arXiv2026

Pb4U-GNet: Resolution-Adaptive Garment Simulation via Propagation-before-Update Graph Network

Garment simulation is fundamental to various applications in computer vision and graphics, from virtual try-on to digital human modelling. However, conventional physics-based methods remain computationally expensive, hindering their application in time-sensitive scenarios. While graph neural networks (GNNs) offer promising acceleration, existing approaches exhibit poor cross-resolution generalisation, demonstrating significant performance degradation on higher-resolution meshes beyond the training distribution. This stems from two key factors: (1) existing GNNs employ fixed message-passing depth that fails to adapt information aggregation to mesh density variation, and (2) vertex-wise displacement magnitudes are inherently resolution-dependent in garment simulation. To address these issues, we introduce Propagation-before-Update Graph Network (Pb4U-GNet), a resolution-adaptive framework that decouples message propagation from feature updates. Pb4U-GNet incorporates two key mechanisms: (1) dynamic propagation depth control, adjusting message-passing iterations based on mesh resolution, and (2) geometry-aware update scaling, which scales predictions according to local mesh characteristics. Extensive experiments show that even trained solely on low-resolution meshes, Pb4U-GNet exhibits strong generalisability across diverse mesh resolutions, addressing a fundamental challenge in neural garment simulation.

cs.CV↗

arXiv2026

XPE and VLT /FORS2 polarimetry challenge the Seyfert-1.9 classification of MCG-05-23-16

We report the third observation of the Seyfert-1.9 active galactic nucleus (AGN) MCG-05-23-16 with the Imaging X-ray Polarimetry Explorer (\textit{IXPE}), together with optical spectro-polarimetry obtained at the Very Large Telescope (VLT), and combined with archival near-ultraviolet, optical and near-infrared polarimetric data. No X-ray polarization was detected in the 2-8 keV band, with a 99\% confidence upper limit of $\leq$2.9\%, further reduced to $\leq$2.5\% when combined with the two past IXPE observations of the same target. Monte Carlo simulations suggest that equatorial coronal models are disfavored if the AGN is indeed a type-1.9/2 AGN, but coronae coplanar to the accretion disk remain consistent if the source is less inclined than previously assumed. \textit{VLT}/FORS2 data reveal a typical type-2 spectrum in total flux, a broad H$\alpha$ line in polarized flux, and strongly wavelength dependent polarization degree and angle, rotating by nearly 70$^\circ$ across the optical band. Comparison with historical measurements confirms long-term stability of the polarization spectrum and a $\sim$90$^\circ$ rotation in the near-ultraviolet. Interpreting the multi-wavelength polarization relative to the AGN ionization axis indicates that the main obscurer is not a compact circumnuclear torus, but a distant kpc-scale dust lane crossing the galaxy. This result implies that MCG-05-23-16 is in fact a type-1 AGN seen through foreground dust. The low X-ray column density becomes consistent with the absence of polarization, provided that the nuclear inclination is low.

astro-ph.HE↗

arXiv2026

Benchmark^2: Systematic Evaluation of LLM Benchmarks

The rapid proliferation of benchmarks for evaluating large language models (LLMs) has created an urgent need for systematic methods to assess benchmark quality itself. We propose Benchmark^2, a comprehensive framework comprising three complementary metrics: (1) Cross-Benchmark Ranking Consistency, measuring whether a benchmark produces model rankings aligned with peer benchmarks; (2) Discriminability Score, quantifying a benchmark's ability to differentiate between models; and (3) Capability Alignment Deviation, identifying problematic instances where stronger models fail but weaker models succeed within the same model family. We conduct extensive experiments across 15 benchmarks spanning mathematics, reasoning, and knowledge domains, evaluating 11 LLMs across four model families. Our analysis reveals significant quality variations among existing benchmarks and demonstrates that selective benchmark construction based on our metrics can achieve comparable evaluation performance with substantially reduced test sets.

cs.CL↗

arXiv2025

Pulsed, Polarized X-ray Emission from Neutron Star Surfaces: the Effects of Vacuum Birefringence in the Magnetosphere

Intense magnetic fields in the atmospheres of neutron stars render non-trivial angular dependence of intensity and polarization of soft X-ray emission originating from their surfaces. By tracking the complex electric field vector for each photon during its atmospheric transport and propagation in general relativistic and birefringent magnetospheres, our Monte Carlo simulation, named MAGTHOMSCATT, allows for capturing the complete polarization properties, including the intricate interplay between linearity and circularity. The new inclusion in MAGTHOMSCATT of quantum electrodynamical influences on polarization in the magnetosphere is presented. We simulate the pulsed and polarized X-ray emission from the outer layers of optically thick, fully ionized atmospheres of neutron stars, with a focus on the radiation emitted from extended polar caps of magnetars, which are the most highly magnetized neutron stars. Using the recent intensity pulse profile data for the magnetar 1RXS J11708-4009, we constrain the geometric parameters, namely the angles between the magnetic axis and the observer's viewing direction relative to the spin axis, as well as the sizes of emission regions. The distributions of these parameters and the best-fit configuration are provided. In addition, we discuss the important impacts of vacuum birefringence in the magnetosphere on increasing the linear polarization degree. A comparison with the case of a weakly magnetized neutron star, RX J0822.0-4300, is also discussed. Our simulation still needs further development, particularly to incorporate the vacuum resonance effect. Nevertheless, the formalism presented here can be employed to constrain geometric parameters for various types of neutron stars.

astro-ph.HE↗

arXiv2025

Multiplexed SiPM Readout of Plastic Scintillating Fiber Detector for Muon Tomography

Muon tomography is a non-destructive imaging technique that uses cosmic-ray muons to probe dense materials. Bar scintillator and scintillating fiber detectors equipped with one-dimensional SiPM arrays offer compact, high-resolution solutions, but large-area implementations require effective reduction of readout channels while preserving detector performance. To address this challenge, we present a novel multiplexing scheme based on a diode-based symmetric charge division circuit combined with a position-encoding algorithm, enabling up to $N_{\textrm{SiPM}}^{\textrm{max}}=C^{2}_{N_{\textrm{ele}}}$ SiPM channels to be read out using only ${N_{\textrm{ele}}}$ electronic channels. Circuit simulations confirm the feasibility of the multiplexing design and guide the choice of appropriate diodes to preserve SiPM signal integrity. The approach was validated using a SciFi detector module comprising 21 SiPM channels multiplexed into 7 electronic channels. Electronic tests show that this multiplexing circuit exhibits low crosstalk between electronic channels, and preserves linearity over a dynamic range from $\sim$10 to 122 photoelectrons. Cosmic-ray measurements further show that the multiplexed SciFi detector achieves a detection efficiency above 95\% and a spatial resolution of about 0.65~mm, with only minor degradation compared to the direct (per SiPM channel) readout. These results verify that the proposed method provides a scalable and cost-effective readout solution for large-area muon tomography systems and is applicable to other scintillator-based detectors employing similar one-dimensional SiPM array readout.

physics.ins-det↗

arXiv2025

CM-LIUW-Odometry: Robust and High-Precision LiDAR-Inertial-UWB-Wheel Odometry for Extreme Degradation Coal Mine Tunnels

Simultaneous Localization and Mapping (SLAM) in large-scale, complex, and GPS-denied underground coal mine environments presents significant challenges. Sensors must contend with abnormal operating conditions: GPS unavailability impedes scene reconstruction and absolute geographic referencing, uneven or slippery terrain degrades wheel odometer accuracy, and long, feature-poor tunnels reduce LiDAR effectiveness. To address these issues, we propose CoalMine-LiDAR-IMU-UWB-Wheel-Odometry (CM-LIUW-Odometry), a multimodal SLAM framework based on the Iterated Error-State Kalman Filter (IESKF). First, LiDAR-inertial odometry is tightly fused with UWB absolute positioning constraints to align the SLAM system with a global coordinate. Next, wheel odometer is integrated through tight coupling, enhanced by nonholonomic constraints (NHC) and vehicle lever arm compensation, to address performance degradation in areas beyond UWB measurement range. Finally, an adaptive motion mode switching mechanism dynamically adjusts the robot's motion mode based on UWB measurement range and environmental degradation levels. Experimental results validate that our method achieves superior accuracy and robustness in real-world underground coal mine scenarios, outperforming state-of-the-art approaches. We open source our code of this work on Github to benefit the robotics community.

cs.RO↗

arXiv2025

Text2Sign Diffusion: A Generative Approach for Gloss-Free Sign Language Production

Sign language production (SLP) aims to translate spoken language sentences into a sequence of pose frames in a sign language, bridging the communication gap and promoting digital inclusion for deaf and hard-of-hearing communities. Existing methods typically rely on gloss, a symbolic representation of sign language words or phrases that serves as an intermediate step in SLP. This limits the flexibility and generalization of SLP, as gloss annotations are often unavailable and language-specific. Therefore, we present a novel diffusion-based generative approach - Text2Sign Diffusion (Text2SignDiff) for gloss-free SLP. Specifically, a gloss-free latent diffusion model is proposed to generate sign language sequences from noisy latent sign codes and spoken text jointly, reducing the potential error accumulation through a non-autoregressive iterative denoising process. We also design a cross-modal signing aligner that learns a shared latent space to bridge visual and textual content in sign and spoken languages. This alignment supports the conditioned diffusion-based process, enabling more accurate and contextually relevant sign language generation without gloss. Extensive experiments on the commonly used PHOENIX14T and How2Sign datasets demonstrate the effectiveness of our method, achieving the state-of-the-art performance.

cs.CL↗

arXiv2025

Unveiling blazar synchrotron emission: a multiwavelength polarimetric study of HSP and LSP populations

Polarimetric properties of blazars allow us to put constraints on the acceleration mechanisms that fuel their powerful jets. By studying the multiwavelength polarimetric behaviour of high synchrotron peaked (HSP) and low synchrotron peaked (LSP) blazars, we aim to explore differences in their emission mechanisms and magnetic field structure in the acceleration region. In this study, we take advantage of several X-ray polarisation observations of HSP by the IXPE, including four new observations of Mrk 501, and optical polarisation observations of LSP from RoboPol and many others. We find that the polarisation degree (PD) distribution of HSP in X-rays is systematically higher than in optical and mm-radio wavelengths, as reported in previous IXPE publications. The distribution of the X-ray electric vector position angles (PA) is centered around the jet axis with most of the observations consistent with zero difference within uncertainties. In fact, the distribution of the offset of the PA from the jet axis is consistent between the LSP and HSP populations (with PA measured in optical for the first, X-ray for the latter), suggesting a common magnetic field structure close to the acceleration region, in strong support of the emerging energy stratified picture of particle acceleration followed by energy loss in blazar jets.

astro-ph.HE↗

arXiv2025

Monte Carlo Simulations of Polarized Radiative Transfer in Neutron Star Atmospheres

Soft X-ray emission from neutron stars affords powerful diagnostic tools for uncovering their surface and interior properties, as well as their geometric configurations. In the atmospheres of neutron stars, the presence of magnetic fields alters the photon-electron scattering cross sections, resulting in non-trivial angular dependence of intensity and polarization of the emergent signals. This paper presents recent developments of our Monte Carlo simulation, MAGTHOMSCATT, which tracks the complex electric field vector for each photon during its transport. Our analysis encompasses the anisotropy and polarization characteristics of X-ray emission for field strengths ranging from non-magnetic to extremely magnetized regimes that are germane to magnetars. In the very low field domain, we reproduced the numerical solution to the radiative transfer equation for non-magnetic Thomson scattering, and provided analytical fits for the angular dependence of the intensity and the polarization degree. These fits can be useful for studies of millisecond pulsars and magnetic white dwarfs. By implementing a refined injection protocol, we show that, in the magnetar regime, the simulated intensity and polarization pulse profiles of emission from extended surface regions becomes invariant with respect to the ratio of photon ($\omega$) and electron cyclotron ($\omega_{\rm B}$) frequencies once $\omega / \omega_{\rm B} \lesssim 0.01$. This circumvents the need for simulations pertinent to really high magnetic field strengths, which are inherently slower. Our approach will be employed elsewhere to model observational data to constrain neutron star geometric parameters and properties of emitting hot spots on their surfaces.

astro-ph.HE↗

arXiv2025

XL-Calibur Polarimetry of Cyg X-1 Further Constrains the Origin of its Hard-state X-ray Emission

The balloon-borne hard X-ray polarimetry mission XL-Calibur observed the Black Hole X-ray Binary (BHXRB) Cygnus X-1 (Cyg X-1) during its nearly six-day Long Duration Balloon (LDB) flight from Sweden to Canada in July 2024. The XL-Calibur observations allowed us to derive the most precise constraints to date of the Polarization Degree (PD) and Polarization Angle (PA) of the hard X-ray emission from a BHXRB. XL-Calibur observed Cyg X-1 in the hard state and measured a $\sim$19-64 keV PD of ($5.0^{+2.7}_{-3.0}$)% at a PA of $-28^{\circ}\pm 17^{\circ}$, with an 8.7% chance probability of detecting larger PDs than the one observed, given an unpolarized signal. The XL-Calibur results are thus comparable to the 2-8 keV PD and PA found by IXPE, with a similar agreement between the hard X-ray PA and the radio jet direction. We also discuss the implications of our polarization measurements in the context of models describing the origin of the broadband X-ray and $\gamma$-ray emission, to which XL-Calibur provides independent constraints on any proposed emission modeling.

astro-ph.HE↗

arXiv2025

PUMPS: Skeleton-Agnostic Point-based Universal Motion Pre-Training for Synthesis in Human Motion Tasks

Motion skeletons drive 3D character animation by transforming bone hierarchies, but differences in proportions or structure make motion data hard to transfer across skeletons, posing challenges for data-driven motion synthesis. Temporal Point Clouds (TPCs) offer an unstructured, cross-compatible motion representation. Though reversible with skeletons, TPCs mainly serve for compatibility, not for direct motion task learning. Doing so would require data synthesis capabilities for the TPC format, which presents unexplored challenges regarding its unique temporal consistency and point identifiability. Therefore, we propose PUMPS, the primordial autoencoder architecture for TPC data. PUMPS independently reduces frame-wise point clouds into sampleable feature vectors, from which a decoder extracts distinct temporal points using latent Gaussian noise vectors as sampling identifiers. We introduce linear assignment-based point pairing to optimise the TPC reconstruction process, and negate the use of expensive point-wise attention mechanisms in the architecture. Using these latent features, we pre-train a motion synthesis model capable of performing motion prediction, transition generation, and keyframe interpolation. For these pre-training tasks, PUMPS performs remarkably well even without native dataset supervision, matching state-of-the-art performance. When fine-tuned for motion denoising or estimation, PUMPS outperforms many respective methods without deviating from its generalist architecture.

cs.CV↗

arXiv2025

Fuzzing: Randomness? Reasoning! Efficient Directed Fuzzing via Large Language Models

Fuzzing is highly effective in detecting bugs due to the key contribution of randomness. However, randomness significantly reduces the efficiency of fuzzing, causing it to cost days or weeks to expose bugs. Even though directed fuzzing reduces randomness by guiding fuzzing towards target buggy locations, the dilemma of randomness still challenges directed fuzzers. Two critical components, which are seeds and mutators, contain randomness and are closely tied to the conditions required for triggering bugs. Therefore, to address the challenge of randomness, we propose to use large language models (LLMs) to remove the randomness in seeds and reduce the randomness in mutators. With their strong reasoning and code generation capabilities, LLMs can be used to generate reachable seeds that target pre-determined locations and to construct bug-specific mutators tailored for specific bugs. We propose RandLuzz, which integrates LLMs and directed fuzzing, to improve the quality of seeds and mutators, resulting in efficient bug exposure. RandLuzz analyzes function call chain or functionality to guide LLMs in generating reachable seeds. To construct bug-specific mutators, RandLuzz uses LLMs to perform bug analysis, obtaining information such as bug causes and mutation suggestions, which further help generate code that performs bug-specific mutations. We evaluate RandLuzz by comparing it with four state-of-the-art directed fuzzers, AFLGo, Beacon, WindRanger, and SelectFuzz. With RandLuzz-generated seeds, the fuzzers achieve an average speedup ranging from 2.1$\times$ to 4.8$\times$ compared to using widely-used initial seeds. Additionally, when evaluated on individual bugs, RandLuzz achieves up to a 2.7$\times$ speedup compared to the second-fastest exposure. On 8 bugs, RandLuzz can even expose them within 60 seconds.

cs.SE↗

arXiv2025

F2Net: A Frequency-Fused Network for Ultra-High Resolution Remote Sensing Segmentation

Semantic segmentation of ultra-high-resolution (UHR) remote sensing imagery is critical for applications like environmental monitoring and urban planning but faces computational and optimization challenges. Conventional methods either lose fine details through downsampling or fragment global context via patch processing. While multi-branch networks address this trade-off, they suffer from computational inefficiency and conflicting gradient dynamics during training. We propose F2Net, a frequency-aware framework that decomposes UHR images into high- and low-frequency components for specialized processing. The high-frequency branch preserves full-resolution structural details, while the low-frequency branch processes downsampled inputs through dual sub-branches capturing short- and long-range dependencies. A Hybrid-Frequency Fusion module integrates these observations, guided by two novel objectives: Cross-Frequency Alignment Loss ensures semantic consistency between frequency components, and Cross-Frequency Balance Loss regulates gradient magnitudes across branches to stabilize training. Evaluated on DeepGlobe and Inria Aerial benchmarks, F2Net achieves state-of-the-art performance with mIoU of 80.22 and 83.39, respectively. Our code will be publicly available.

cs.CV↗

arXiv2025

The Cygnus X-1 Puzzle: Implications of X-ray Polarization Measurements in the Soft and Hard States on the Properties of the Accretion Flow and the Emission Mechanisms

In this paper, we summarize key observational constraints of the accretion flow on the black hole X-ray binary Cygnus X-1 (Cyg X-1). The discussion highlights the flows of energy close to the black hole and the importance of the distance range from which the radiating zone draws its energy. For the hard state, we examine compact and extended corona models. We find that compact corona models are energetically favored, but extended models cannot be fully excluded. We discuss the high linear polarization of the Cyg X-1 X-rays in the soft and hard states, parallel to the direction of the radio jet. We propose the presence of a pair layer enveloping the accretion disk moving at approximately half the speed of light away from the disk for both the soft and the hard state. In the soft state, the pairs cool to the Compton temperature of the disk emission. In the hard state, the pairs acquire thermal and bulk motion allowing them to Comptonize the emission to produce the observed power law emission. In both emission states, the bulk motion away from the disk leads to a net polarization parallel to the radio jet. We emphasize that the geometry of the accretion flow in the hard state is still not well constrained, and that observed spectral (including the relativistically broadened Fe K-$\alpha$ line) and spectro-polarimetric signatures depend strongly on the plasma processes responsible for energy dissipation in the plasma.

astro-ph.HE↗