arXiv subjects

Jun Yang

Publications and source records attributed to Jun Yang.

At least 55 records · Page 3Linked to original sources

arXiv2026

ProMSA:Progressive Multimodal Search Agents for Knowledge-Based Visual Question Answering

Knowledge-based Visual Question Answering (KB-VQA) requires models to combine image understanding with external knowledge. Most prior methods use a fixed retrieve-then-generate pipeline with a pre-selected retriever and a static top-k setting, which is not adaptive during reasoning. We propose ProMSA, a progressive multimodal search agent for KB-VQA. Given an image-question pair, the agent iteratively chooses image search, text search, or stop, under explicit tool-call budgets and with deduplication to avoid redundant retrieval. For training, we first use rejection-sampling SFT to learn valid tool-use formats, then optimize the agent with TN-GSPO, a sequence-level RL objective that normalizes updates by both generation length and tool-interaction depth. Experiments on E-VQA and InfoSeek show consistent gains over strong RAG and agent baselines, and improved retrieval and end-to-end accuracy. The code is available at https://github.com/DingWu1021/Promsa.

cs.CV↗

arXiv2026

Exploring Tidal Disruption Events with SKA and VLBI: Unveiling the Mystery of Black Hole Feeding and Outflows

Tidal disruption events (TDEs) probe the birth and evolution of black hole accretion flows and jets on human timescales. Radio emission traces shocks and outflows from thermal TDEs and powerful relativistic jets in the rare jetted class. SKA Mid, phased for VLBI and used together with global networks, will deliver milliarcsecond imaging, tens of microarcsecond astrometry, and microJy sensitivity, enabling: (i) proper motion measurements that discriminate off axis relativistic jets from subrelativistic winds; (ii) resolved morphologies and magnetic field diagnostics via polarimetry; and (iii) precise nuclear localization to distinguish SMBH vs. IMBH and to reveal recoiling or binary systems. SKA's wide frequency coverage (0.35 to 15.4 GHz) and 1h continuum sensitivities of 3 to 10 microJy per beam, together with multibeam tiedarray VLBI and a transient buffer for rapid triggers, are transformational. LSST, Einstein Probe, and SVOM will increase TDE alerts to hundreds per year, and late time radio flares appear common, ensuring rich SKA VLBI samples. We provide observing strategies, detection forecasts, and predictions, e.g., about 5 proper motion detections of jetted (or off axis) TDEs per year and routine core shift constraints at the microarcsecond level. This program will establish TDEs as laboratories for exploring jet launching, particle acceleration (including neutrinos), black hole accretion history and demographics, and properties of circumnuclear medium.

astro-ph.HE↗

arXiv2026

AGN Jets from Formation to Dissipation

Active Galactic Nuclei (AGN) are among the most energetic phenomena in the Universe, capable of launching powerful relativistic jets that extend from sub-parsec to megaparsec scales. These jets play a crucial role in regulating star formation, redistributing energy and matter, and shaping the evolution of galaxies and their environments. Despite decades of study, a comprehensive understanding of how AGN jets form, propagate, and dissipate remains elusive. The aim of this chapter is to highlight how the future capabilities of the the Square Kilometre Array (SKA), as a standalone array as well as in combination with Very Long Baseline Interferometry (VLBI) arrays and multi-wavelength facilities, will transform our capabilities to study the co-evolution of AGN jets and their host galaxies from jet formation to dissipation scales.

astro-ph.GA↗

arXiv2026

Evaluating Learned Spatial Indexes

Learned indexes improve query performance by adapting search structures to data and workload distributions. Although many learned indexes have been proposed, their trade-offs remain insufficiently understood for spatial range queries, where performance depends not only on model accuracy but also on data and query skew, layout granularity, selectivity, and storage behavior. In this work, we perform an experimental study of learned indexes for spatial range queries. We examine a representative set of indexes and address seven fundamental questions: (1) How does block size influence query latency, and what configurations yield optimal performance under varying selectivities? (2) How do skewed data and query distributions impact index performance? (3) How do indexes balance refinement and scan costs, and which designs favor one over the other? (4) How do disk-based storage conditions alter optimal block size and latency trade-offs compared to in-memory settings? (5) What are the construction costs of different indexes, and under what query volumes are these costs amortized? (6) For a given data and query workload, which index is expected to perform best? (7) Do index-selection insights learned from synthetic data generalize to real-world data distributions? To enable the analysis, we use a framework with a common storage backend, standardized query execution pipelines, and controlled variations in data and query skew. Our experiments reveal critical insights into refinement vs. scan trade-offs, the impact of block size, and the interplay between selectivity and layout effectiveness. We synthesize these findings into a workload-based decision tree for index selection and validate it on real OpenStreetMap point sets with synthetic queries, confirming that its recommendations exhibit minimal decision regret and typically yield near-optimal query performance.

cs.DB↗

arXiv2026

SEAGym: An Evaluation Environment for Self-Evolving LLM Agents

Self-evolving LLM-based agents improve mainly by changing their agent harness: the structured execution layer around a base model, including prompts, memory, tools, middleware, runtime state, and the model-tool interaction loop. Existing evaluations often reduce this process to isolated task scores or a single sequential curve, obscuring whether an update produces reusable improvement, overfits recent tasks, increases cost, or harms older behavior. We introduce SEAGym, an evaluation environment for measuring agent harness updates across training, validation, test, replay, and cost records. SEAGym turns Harbor-compatible benchmarks into dynamic self-evolution task sources with train batches, frozen update-validation, held-out ID and OOD transfer views, replay diagnostics, and saved snapshot and metric records. Instantiating SEAGym on Terminal-Bench 2.0 and HLE, we compare ACE, TF-GRPO, and AHE under a shared epoch/batch protocol. The results show that these evaluation views provide complementary signals about the evolution process: frequent updates may fail to improve held-out performance, useful intermediate snapshots may collapse later, and source diversity and model backend can affect harness reliability.

cs.AI↗

arXiv2026

A Jet from a Nearly Dormant Black Hole

Most galaxies host supermassive black holes (SMBHs) that remain weakly accreting or dormant for much of their lifetimes. At the lowest accretion rates, these systems may represent the transition between active nuclei and dormant black holes, but whether they can still launch collimated jets remains unclear. The nuclei in our Galaxy (\sgra) and M31 are key examples of this regime, although no clear jet structure has yet been detected in either source. Here we report multi-frequency very long baseline interferometric observations of \Msixty\ (NGC~4649), a nearby elliptical galaxy hosting a nearly dormant SMBH with an Eddington ratio of $\sim10^{-8}$. We detect a compact two-sided jet with an unusually steep synchrotron spectrum, demonstrating that collimated outflows can persist even under nearly dormant accretion conditions. The apparent radio core exhibits an unprecedentedly steep frequency-dependent position shift toward the SMBH, locating the central engine only $\sim57\,μ$as, corresponding to a projected distance of $\sim10$ Schwarzschild radii, upstream of the 8.37-GHz core. The observed jet morphology and steep core-shift behaviour are reproduced by general relativistic magnetohydrodynamic and radiative-transfer simulations, indicating a magnetically dominated, non-equipartition jet-launching region that departs from the standard conical equipartition picture. These results provide direct observational evidence that jet production can survive near the dormant SMBHs and establish \Msixty\ as a unique laboratory for probing jet formation on event-horizon scales in the lowest-accretion SMBH regime.

astro-ph.HE↗

arXiv2026

Detection of a parsec-scale, compact, and fading ejecta from an accreting massive black hole

Dwarf galaxies, characterized by their low luminosities and masses, are excellent candidates for searches for intermediate-mass black holes (IMBHs), particularly when they show strong accretion and ejection activity. The dwarf galaxy SDSS J101747.09+393207.7 has recently been found to display a very high X-ray luminosity and an X-shaped optical structure, possibly caused by a dwarf--dwarf merger. To explore its potential IMBH ejection activity, we performed very long baseline interferometry (VLBI) observations at 4.9 GHz. In this work, we present the detection of a milliarcsecond-scale, compact, sub-microjansky radio component near the optical centroid. According to some existing radio sky survey data, the radio component was not detected until 2015; it displayed an optically thin steep radio spectrum and declining flux densities across 0.8--5 GHz from 2019 to 2025. Therefore, we identify it as a short-lived and rarely seen ejecta that was produced by unstable accretion onto a massive black hole and likely faded away in a few decades. These results indicate that short-lived, episodic jet activity from accreting IMBHs in dwarf galaxies might exist.

astro-ph.HE↗

arXiv2026

Multi-GPU MBE(3)-OSV-MP2 for Performant Large-Scale ab initio Calculations

The computational acceleration of orbital-invariant local correlation methods on graphics processing units (GPUs) has remained largely unexplored due to substantial algorithmic complexities. The runtime efficiency of GPU-implemented local correlation theories can be significantly constrained by the parallelizable degree of the orbital localization procedure, the iterative solution of the local wave function, and the adaptation of CUDA kernels to inherently local or sparse operations. Using the second-order Møller-Plesset perturbation (MP2) theory, we present a multi-GPU implementation for large-scale third-order many-body expansion orbital-specific virtual MP2 (MBE(3)-OSV-MP2) energy calculations. Accordingly, our algorithms and implementation address the GPU parallelization ability for peak utilization and parallelism of local MP2 computation in several aspects, including Jacobi-Pipek-Mezey localization, randomized OSV generation, direct MP2 integral regeneration, as well as CUDA kernel adaptation to local operations. The GPU-based MBE(3)-OSV-MP2 energy computation achieves $O(N^{1.9})$ scaling and 84\% parallel efficiency up to 24 GPUs distributed on multiple nodes. The present implementation delivers 40-fold wall-time speedup of the canonical RI-MP2 and 10-fold speedup of the CPU-based MBE(3)-OSV-MP2 for (H$_2$O)$_{128}$/cc-pVDZ and (H$_2$O)$_{190}$/cc-pVDZ, respectively. A large scale computation of 784-atom insulin peptide yields the full MBE(3)-OSV-MP2 energies in 24 minutes with cc-pVDZ (7571 basis functions) and 6.4 hours with cc-pVTZ (17448 basis functions) on 8 NVIDIA A800 GPUs. Our work opens up new possibilities for performing fast GPU-based local correlation calculations on real-life macromolecules.

physics.chem-ph↗

arXiv2026

A 4.5-s Quasiperiodic Spectral Oscillation in GRB 230307A: Evidence for Free Precession of a Post-Merger Magnetar?

Millisecond magnetars, rapidly rotating neutron stars with ultra-strong magnetic fields, have long been proposed as central engines of gamma-ray bursts (GRBs). For GRBs produced by neutron star mergers, the survival of a long-lived magnetar remnant remains uncertain, as the merger remnant may rapidly collapse into a black hole. In GRB 230307A, multiwavelength observations together with a previously reported 909-Hz periodic signal consistent with millisecond spin in its prompt emission provide strong evidence that such a post-merger magnetar may power the burst. Here we report the discovery of a quasiperiodic modulation with a characteristic period of 4.5 s in the spectral evolution of GRB 230307A, detected consistently across multiple gamma-ray instruments. The modulation is manifested as a coherent, energy-dependent variation of the spectral shape, with the strongest signature in the evolution of the peak energy. Within the magnetar-engine framework, such a low-frequency modulation can be interpreted as a manifestation of large-scale periodic variations associated with the central engine. If interpreted in terms of free precession, the observed timescale implies a stellar ellipticity of $ε\gtrsim 2.4 \times 10^{-4}$, corresponding to an internal magnetic field strength of $B_t \gtrsim 1.6 \times 10^{16}$ G, alongside a dipole field of $B_p \approx 5.6 \times 10^{15}$ G inferred from the early X-ray emission. These results suggest that such systems may provide potential sources of post-merger gravitational waves (GWs), motivating targeted searches following GRB triggers.

astro-ph.HE↗

arXiv2026

Jet-Structure Imprint on the Curvature Tail of Gamma-Ray Burst Prompt Emission

Even though the prompt emission of gamma-ray bursts (GRBs) is highly beamed, high-latitude emission still produces a distinct light curve break after the intrinsic emission ceases and the edge of the jet comes into view. This curvature effect offers a direct probe of the jet structure during the prompt phase. To uncover the geometric structure of the GRB jet encoded in the prompt light-curve evolution, we develop a numerical model that calculates synchrotron light curves from structured jets to interpret the observed break. We apply this model to the prompt emission of GRB 230307A, which displays a rare late-time break. Our analysis demonstrates that simple spherical outflow and top-hat jet models are inadequate to reproduce the light curve. Instead, the observations are best described by a power-law wing jet with a uniform core ($θ_{\rm core}=0.0147$ rad) and a surrounding power-law wing. Our results demonstrate that the break in late-time prompt emission can be a powerful diagnostic of GRB jet structure.

astro-ph.HE↗

arXiv2026

MAAS-SFRThelper: An Integrated ESAPI Plugin for Structure Generation, Optimization, and Evaluation of Spatially Fractionated Radiation Therapy

Spatially fractionated radiation therapy (SFRT) planning requires three coordinated tasks: generation of high-dose sphere structures, position-aware optimization, and peak-valley dose ratio evaluation. We present MAAS-SFRThelper, a shared-source Eclipse Scripting Application Programming Interface (ESAPI) plugin that integrates structure generation, geometric-aware optimization, and peak-valley dose ratio evaluation for SFRT into a single workflow inside Varian's Eclipse treatment planning system. The plugin exposes five task-oriented tabs sharing common services for sphere extraction and objective creation. The SphereLattice tab generates sphere lattices using five placement patterns. The Optimization tab searches over candidate lattice positions using a four-metric geometric surrogate score and triggers VMAT optimization and dose calculation. The Evaluation tab implements four analysis modes; its three-dimensional peak-valley classification recovers sphere centers from the lattice structure through a geometric extraction pipeline rather than relying on dose thresholds. We validated all functionality on digital phantoms against analytic ground truth. The plugin is distributed as source code under the Varian Limited Use Software License Agreement. Source code and documentation are publicly available on GitHub.

physics.med-ph↗

arXiv2026

SI-Diff: A Framework for Learning Search and High-Precision Insertion with a Force-Domain Diffusion Policy

Contact-rich assembly is fundamental in robotics but poses significant challenges due to uncertainties in relative poses, such as misalignments and small clearances in peg-in-hole tasks. Existing approaches typically address search and high-precision insertion separately, because these tasks involve distinct action patterns. However, supporting both tasks within a single model, without switching models or weights, is desirable for intelligent assembly systems. In this work, we propose SI-Diff, a framework that learns both search and high-precision insertion through a force-domain diffusion policy. To this end, we introduce a new mode-conditioning mechanism that enables the policy to capture distinct action behaviors under a single framework. Moreover, we develop a new search teacher policy that can generate diverse trajectories. By training on successful and efficient demonstrations provided by the teacher policy, the model learns the mapping from tactile and end-effector velocity observations to effective action behaviors. We conduct thorough experiments to show that SI-Diff extends the tolerance to x-y misalignments from 2 mm to 5 mm compared to the state-of-the-art baseline, TacDiffusion, while also demonstrating strong zero-shot transferability to unseen shapes.

cs.RO↗

arXiv2026

Metropolis-Adjusted Diffusion Models

Sampling from score-based diffusion models incurs bias due to both time discretisation and the approximation of the score function. A common strategy for reducing this bias is to apply corrector steps based on the unadjusted Langevin algorithm (ULA) at each noise level within a predictor-corrector framework. However, ULA is itself a biased sampler, as it discretises a continuous diffusion process. In this work, we consider adjusted Langevin correctors that employ Metropolis--Hastings (MH) or Barker's accept-reject steps to correct for this bias. Since the target density ratio typically required by MH-based algorithms is unavailable, we propose methods that instead utilise the score function to compute the correct acceptance probability. We introduce the first exact method for adjusting Langevin corrections in diffusion models, based on a two-coin Bernoulli factory algorithm. We also propose an efficient approximation based on Simpson's rule that achieves accuracy of order $5/2$ in the step size at near-zero marginal cost. We demonstrate that these procedures improve sample quality on both synthetic and image datasets, yielding consistent gains in Fréchet Inception Distance (FID) on the latter.

stat.ML↗

arXiv2026

Stereographic Multiple-Try Metropolis

Multiple-proposal MCMC algorithms have recently gained attention for their potential to improve performance, especially through parallel implementation on modern hardware. We introduce Stereographic Multiple-Try Metropolis (SMTM), a novel family of gradient-free algorithms designed for sampling high-dimensional distributions. By integrating multiple-try Metropolis (MTM) with the stereographic MCMC framework, SMTM overcomes the traditional limitations of MTM, particularly its pathological convergence behavior often observed in high dimensions. For both light-tailed and heavy-tailed targets, SMTM not only outperforms classical MTM and the existing stereographic random-walk Metropolis but also demonstrates strong robustness to tuning. These advantages are supported by high-dimensional scaling analysis and validated through extensive simulation studies.

stat.CO↗

arXiv2026

Detections of nearly bias-free core shifts with 5-30 $μ$as precisions at 8-43 GHz in BL Lacertae

When a radio jet is partially optically thick in the launching region, its apparent compact core may display frequency-dependent positional shifts. High-precision astrometric measurements of core shifts enable astronomers to pinpoint the jet's origin and place tight constraints on the magnetic field. BL Lacertae, the archetypal BL Lac object, hosts a highly variable and well-collimated jet. To independently constrain its innermost core shifts, we conducted very long baseline interferometric (VLBI) observations at 8.4, 12.4, 15.2, 23.6, and 43.2 GHz. By exploiting a nearby (13.3 arcmin) steep-spectrum calibrator (NVSS J220340+420839) through inverse phase-referencing VLBI astrometry, we detect nearly unbiased two-dimensional core shift measurements with state-of-the-art precisions of 5-30 $μ$as, which are significant at $>3σ$ confidence. The core shift between 8.4 and 43.2 GHz reaches 250 $μ$as. The apparent core shifts scale with frequency as $ν^{-1/k_r}$, implying the existence of an optically thick region in the upstream of jet. The derived core-shift index, $k_r\!=\!1.18^{+0.59}_{-0.34}$, is consistent, within uncertainties, with the canonical $k_r\!=\!1$ expected under energy equipartition between the jet particle and magnetic field energy densities, while allowing for modest deviations given that BL Lacertae was captured in a flaring state.

astro-ph.HE↗

arXiv2026

Deep VLBI constraints on compact radio cores in four ultraluminous X-ray sources

We present high-sensitivity Very Long Baseline Interferometry (VLBI) observations of four ultraluminous X-ray sources (ULXs): Holmberg II X-1, IC 342 X-1, NGC 6946 X-1, and NGC 925 X-1. No compact emission was detected on milliarcsecond scales, with rms noise levels reaching approximately 5--20 $μ$Jy. The corresponding $5σ$ flux density upper limits reach $\sim 26\,μ\mathrm{Jy}$, implying radio luminosity limits $L_{\rm R} \lesssim 2 \times 10^{33}\,\mathrm{erg\,s^{-1}}$. This disfavors any persistently bright hard-state-like compact core at our sensitivity level. The previously reported VLBI core in Holmberg II X-1 exhibits significant long-term variability, broadly consistent with an overall decline over the past decades. This behavior is consistent with emission from optically-thin ejecta undergoing adiabatic expansion. The VLBI non-detections may reflect intrinsically weak/intermittent compact emission, and/or low--surface--brightness structure that is resolved out by VLBI, and/or absorption/propagation effects such as free--free absorption in dense, ionized winds.

astro-ph.HE↗

arXiv2026

Three-dimensional transport-induced chemistry on temperate sub-Neptune K2-18b, Part II: the combined effects of atmospheric dynamics and chemical reactions

The upper atmospheres of temperate sub-Neptunes are strongly influenced by atmospheric dynamics due to their cool equilibrium temperature and thereby longer chemical timescales than the atmospheric dynamical timescales. In this study, we used a three-dimensional (3D) general circulation model to investigate the transport-induced disequilibrium chemistry and vertical mixing on temperate gas-rich mini-Neptunes, using K2-18b as an example. We model K2-18b assuming 180 times solar metallicity and consider it as either a synchronous or an asynchronous rotator, exploring spin-orbit resonances of 2:1, 6:1, and 10:1. We find that the vertical transport affects the chemical structure significantly, making CO$_2$ and CO more abundant ($\sim$10$^{-3}$) in the upper atmosphere compared to the chemical equilibrium abundance (<10$^{-15}$), and horizontal winds further homogenize the chemical composition zonally in this region. Molecular abundances in the photosphere generally agree across different rotation periods. We employ a passive tracer in the model to estimate the one-dimensional (1D) equivalent eddy-diffusion coefficient ($K_{zz}$) of K2-18b, providing a parameter useful for future 1D atmospheric models. Additionally, synthetic transmission spectra generated from our model are compared with the JWST observations, and we find that our model can provide a comparable fit to the observations. This work offers a 3D perspective on transport-induced chemistry on a temperate sub-Neptune and derives vertical mixing parameters to support 1D modelling.

astro-ph.EP↗

arXiv2026

An Arithmetic Invariant of the Jacquet-Langlands correspondence

We describe the local-global compatibility of local Plancherel measures and the Tamagawa measure under the Jacquet-Langlands correspondence. We apply the notion of densities of modules over a discrete group, which generalizes the dimensions over a discrete group. We prove that the global Jacquet-Langlands correspondence preserves the densities over principal arithmetic groups.

math.NT↗