arXiv subjects

Siqi Zhang

Publications and source records attributed to Siqi Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Distance Estimation and Sky Localization of Eccentric Double White Dwarf Binaries from Gravitational Wave Observations inside Globular Clusters

The cosmic distance scale is built on multiple different techniques for estimating distances in space that are often connected and dependent on multiple measurements and assumptions. Double white dwarf binaries (DWDs) are common objects and are expected to produce gravitational wave (GW) signals that can be observed with space-based detectors such as LISA. By analyzing these signals we should be able to estimate the distance and sky location of the source. Previous studies have done this for circular binaries which, while they are abundant, have, in general, weaker signals than eccentric binaries and it is not possible to differentiate whether a circular binary is in the field or in a dense environment such as a globular cluster (GC). In this paper we used eccentric binaries from MOCCA GC simulations, simulated the GW signal from each binary at locations related to GCs in the Milky Way and estimated the precision on the distance and the sky location of the source. We find that distances can be estimated with higher precision than current day methods even with low eccentricity binaries and higher eccentricity further increases this precision. Although the probability of finding a tight and eccentric DWD is far lower than a circular one, we can expect to find at least a few in the dense environments of the Milky Way, such as GCs. These estimations would be independent measurements with high precision to objects inside dense environments, such as GCs inside the Milky Way and the Magellanic Clouds.

astro-ph.HE↗

arXiv2026

Canzona: A Unified, Asynchronous, and Load-Balanced Framework for Distributed Matrix-based Optimizers

The scaling of Large Language Models (LLMs) drives interest in matrix-based optimizers (e.g., Shampoo, Muon, SOAP) for their convergence efficiency; yet their requirement for holistic updates conflicts with the tensor fragmentation in distributed frameworks like Megatron. Existing solutions are suboptimal: synchronous approaches suffer from computational redundancy, while layer-wise partitioning fails to reconcile this conflict without violating the geometric constraints of efficient communication primitives. To bridge this gap, we propose Canzona, a Unified, Asynchronous, and Load-Balanced framework that decouples logical optimizer assignment from physical parameter distribution. For Data Parallelism, we introduce an alpha-Balanced Static Partitioning strategy that respects atomicity while neutralizing the load imbalance. For Tensor Parallelism, we design an Asynchronous Compute pipeline utilizing Micro-Group Scheduling to batch fragmented updates and hide reconstruction overhead. Extensive evaluations on the Qwen3 model family (up to 32B parameters) on 256 GPUs demonstrate that our approach preserves the efficiency of established parallel architectures, achieving a 1.57x speedup in end-to-end iteration time and reducing optimizer step latency by 5.8x compared to the baseline.

cs.DC↗

arXiv2026

A Unified View of Attention and Residual Sinks: Outlier-Driven Rescaling is Essential for Transformer Training

We investigate the functional role of emergent outliers in large language models, specifically attention sinks (a few tokens that consistently receive large attention logits) and residual sinks (a few fixed dimensions with persistently large activations across most tokens). We hypothesize that these outliers, in conjunction with the corresponding normalizations (\textit{e.g.}, softmax attention and RMSNorm), effectively rescale other non-outlier components. We term this phenomenon \textit{outlier-driven rescaling} and validate this hypothesis across different model architectures and training token counts. This view unifies the origin and mitigation of both sink types. Our main conclusions and observations include: (1) Outliers function jointly with normalization: removing normalization eliminates the corresponding outliers but degrades training stability and performance; directly clipping outliers while retaining normalization leads to degradation, indicating that outlier-driven rescaling contributes to training stability. (2) Outliers serve more as rescale factors rather than contributors, as the final contributions of attention and residual sinks are significantly smaller than those of non-outliers. (3) Outliers can be absorbed into learnable parameters or mitigated via explicit gated rescaling, leading to improved training performance (average gain of 2 points) and enhanced quantization robustness (1.2 points degradation under W4A4 quantization).

cs.CL↗

arXiv2025

How Mass Flows Through Accretion Discs: A Spectral-Timing Vision for the 2040s

Understanding how mass and angular momentum flow through accretion discs remains a fundamental unsolved problem in astrophysics. Accreting white dwarfs offer an ideal laboratory for addressing this question: their variability occurs on accessible timescales of seconds to minutes, and their optical spectra contain continuum and emission-line components that trace distinct disc regions. Broad-band timing studies have revealed time-lags similar to those observed in X-ray binaries and active galactic nuclei, suggesting propagating fluctuations and possible coupling to an inner hot flow. However, the blending of line and continuum light in broad filters prevents a physical interpretation of these signals. The 2040s will bring an unprecedented number of disc-accreting systems discovered by Rubin-LSST, space-based gravitational-wave observatories, and third-generation ground and space-based detectors. To extract disc physics from these sources, high-cadence optical spectral-timing, simultaneously resolving continuum and individual lines, is essential. Such measurements would directly map how variability propagates through discs, determine how the outer disc responds to changes in the inner flow, and test whether accretion physics is scale-invariant from white dwarfs to supermassive black holes. This white paper outlines the scientific motivation and observational capabilities required to realise this vision. It highlights the opportunity for ESO to enable a transformative new window on accretion physics in the coming decade.

astro-ph.IM↗

arXiv2025

Can GNNs Learn Link Heuristics? A Concise Review and Evaluation of Link Prediction Methods

This paper explores the ability of Graph Neural Networks (GNNs) in learning various forms of information for link prediction, alongside a brief review of existing link prediction methods. Our analysis reveals that GNNs cannot effectively learn structural information related to the number of common neighbors between two nodes, primarily due to the nature of set-based pooling of the neighborhood aggregation scheme. Also, our extensive experiments indicate that trainable node embeddings can improve the performance of GNN-based link prediction models. Importantly, we observe that the denser the graph, the greater such the improvement. We attribute this to the characteristics of node embeddings, where the link state of each link sample could be encoded into the embeddings of nodes that are involved in the neighborhood aggregation of the two nodes in that link sample. In denser graphs, every node could have more opportunities to attend the neighborhood aggregation of other nodes and encode states of more link samples to its embedding, thus learning better node embeddings for link prediction. Lastly, we demonstrate that the insights gained from our research carry important implications in identifying the limitations of existing link prediction methods, which could guide the future development of more robust algorithms.

cs.SI↗

arXiv2025

CO-QLink: Cryogenic Optical Link for Scalable Quantum Computing Systems and High-Performance Cryogenic Computing Systems

Cryogenic systems necessitate extensive data transmission between room-temperature and cryogenic environments, as well as within the cryogenic temperature domain. High-speed, low-power data transmission is pivotal to enabling the deployment of larger-scale cryogenic systems, including the scalable quantum computing systems and the high-performance cryogenic computing systems fully immersed in liquid nitrogen. In contrast to wireline and microwave links, optical communication links are emerging as a solution characterized by high data rates, high energy efficiency, low signal attenuation, absence of thermal conduction, and superior scalability. In this work, a 4K heat-insulated high-speed (56Gbps) low-power (1.6pJ/b) transceiver (TRX) that achieves a complete link between 4K systems and room temperature (RT) equipment is presented. Copackaged with a PIN photodiode (PD), the RX uses an inverter-based analog front-end and an analog half-rate clock data recovery loop. Connecting to a Mach-Zehnder modulator (MZM), the TX contains a voltage-mode driver with current-mode injection for low-power output-swing-boosting and 3-tap feed-forward equalization (FFE). This link has been demonstrated in the control and readout of a complete superconducting quantum computing system.

quant-ph↗

arXiv2025

GeoLink: Empowering Remote Sensing Foundation Model with OpenStreetMap Data

Integrating ground-level geospatial data with rich geographic context, like OpenStreetMap (OSM), into remote sensing (RS) foundation models (FMs) is essential for advancing geospatial intelligence and supporting a broad spectrum of tasks. However, modality gap between RS and OSM data, including differences in data structure, content, and spatial granularity, makes effective synergy highly challenging, and most existing RS FMs focus on imagery alone. To this end, this study presents GeoLink, a multimodal framework that leverages OSM data to enhance RS FM during both the pretraining and downstream task stages. Specifically, GeoLink enhances RS self-supervised pretraining using multi-granularity learning signals derived from OSM data, guided by cross-modal spatial correlations for information interaction and collaboration. It also introduces image mask-reconstruction to enable sparse input for efficient pretraining. For downstream tasks, GeoLink generates both unimodal and multimodal fine-grained encodings to support a wide range of applications, from common RS interpretation tasks like land cover classification to more comprehensive geographic tasks like urban function zone mapping. Extensive experiments show that incorporating OSM data during pretraining enhances the performance of the RS image encoder, while fusing RS and OSM data in downstream tasks improves the FM's adaptability to complex geographic scenarios. These results underscore the potential of multimodal synergy in advancing high-level geospatial artificial intelligence. Moreover, we find that spatial correlation plays a crucial role in enabling effective multimodal geospatial data integration. Code, checkpoints, and using examples are released at https://github.com/bailubin/GeoLink_NeurIPS2025

cs.CV↗

arXiv2025

LightCom: A Generative AI-Augmented Framework for QoE-Oriented Communications

Data-intensive and immersive applications, such as virtual reality, impose stringent quality of experience (QoE) requirements that challenge traditional quality of service (QoS)-driven communication systems. This paper presents LightCom, a lightweight encoding and generative AI (GenAI)-augmented decoding framework, designed for QoE-oriented communications under low signal-to-noise ratio (SNR) conditions. LightCom simplifies transmitter design by applying basic low-pass filtering for source coding and minimal channel coding, significantly reducing processing complexity and energy consumption. At the receiver, GenAI models reconstruct high-fidelity content from highly compressed and degraded signals by leveraging generative priors to infer semantic and structural information beyond traditional decoding capabilities. The key design principles are analyzed, along with the sufficiency and error-resilience of the source representation. We also develop importance-aware power allocation strategies to enhance QoE and extend perceived coverage. Simulation results demonstrate that LightCom achieves up to a $14$ dB improvement in robustness and a $9$ dB gain in perceived coverage, outperforming traditional QoS-driven systems relying on sophisticated source and channel coding. This paradigm shift moves communication systems towards human-centric QoE metrics rather than bit-level fidelity, paving the way for more efficient and resilient wireless networks.

eess.SP↗

arXiv2025

Formation Channels of Gravitationally Resolvable Double White Dwarf Binaries Inside Globular Clusters

Current gravitational wave detectors are sensitive to coalescing black holes and neutron stars. However, double white dwarfs (DWDs) have long been recognized as promising sources of gravitational waves, and upcoming detectors like LISA will be able to observe these systems in abundance. DWDs are expected to be the dominant gravitational wave (GW) sources in parts of the LISA frequency range, making it crucial to understand their formation for future detections. The Milky Way contains many white dwarfs (WDs) in both the field and star clusters, promising a rich population of DWDs for LISA. However, the large number of sources may make it difficult to resolve individual binaries, and DWDs in the field and clusters often have similar properties, complicating the identification of their origins from GW signals alone. In this work, we focus on eccentric and tight DWDs, which cannot form in the field, but require dynamical interactions in dense clusters to increase their eccentricity after circularization through mass transfer phases and common-envelope evolution during binary evolution. These binaries may also form in three- and four-body dynamical interactions where a DWD binary may directly form with high eccentricity and low separation. Our results show that we should expect eccentric and tight DWDs in clusters that can provide meaningful GW signal, however, the number is low; with an upper limit of 10-15 in the MW. These can be used to independently obtain distances of their host cluster.

astro-ph.SR↗

arXiv2025

NavBench: Probing Multimodal Large Language Models for Embodied Navigation

Multimodal Large Language Models (MLLMs) have demonstrated strong generalization in vision-language tasks, yet their ability to understand and act within embodied environments remains underexplored. We present NavBench, a benchmark to evaluate the embodied navigation capabilities of MLLMs under zero-shot settings. NavBench consists of two components: (1) navigation comprehension, assessed through three cognitively grounded tasks including global instruction alignment, temporal progress estimation, and local observation-action reasoning, covering 3,200 question-answer pairs; and (2) step-by-step execution in 432 episodes across 72 indoor scenes, stratified by spatial, cognitive, and execution complexity. To support real-world deployment, we introduce a pipeline that converts MLLMs' outputs into robotic actions. We evaluate both proprietary and open-source models, finding that GPT-4o performs well across tasks, while lighter open-source models succeed in simpler cases. Results also show that models with higher comprehension scores tend to achieve better execution performance. Providing map-based context improves decision accuracy, especially in medium-difficulty scenarios. However, most models struggle with temporal understanding, particularly in estimating progress during navigation, which may pose a key challenge.

cs.CV↗

arXiv2025

MedSG-Bench: A Benchmark for Medical Image Sequences Grounding

Visual grounding is essential for precise perception and reasoning in multimodal large language models (MLLMs), especially in medical imaging domains. While existing medical visual grounding benchmarks primarily focus on single-image scenarios, real-world clinical applications often involve sequential images, where accurate lesion localization across different modalities and temporal tracking of disease progression (e.g., pre- vs. post-treatment comparison) require fine-grained cross-image semantic alignment and context-aware reasoning. To remedy the underrepresentation of image sequences in existing medical visual grounding benchmarks, we propose MedSG-Bench, the first benchmark tailored for Medical Image Sequences Grounding. It comprises eight VQA-style tasks, formulated into two paradigms of the grounding tasks, including 1) Image Difference Grounding, which focuses on detecting change regions across images, and 2) Image Consistency Grounding, which emphasizes detection of consistent or shared semantics across sequential images. MedSG-Bench covers 76 public datasets, 10 medical imaging modalities, and a wide spectrum of anatomical structures and diseases, totaling 9,630 question-answer pairs. We benchmark both general-purpose MLLMs (e.g., Qwen2.5-VL) and medical-domain specialized MLLMs (e.g., HuatuoGPT-vision), observing that even the advanced models exhibit substantial limitations in medical sequential grounding tasks. To advance this field, we construct MedSG-188K, a large-scale instruction-tuning dataset tailored for sequential visual grounding, and further develop MedSeq-Grounder, an MLLM designed to facilitate future research on fine-grained understanding across medical sequential images. The benchmark, dataset, and model are available at https://huggingface.co/MedSG-Bench

cs.CV↗

arXiv2025

Enhanced Adaptive Gradient Algorithms for Nonconvex-PL Minimax Optimization

Minimax optimization recently is widely applied in many machine learning tasks such as generative adversarial networks, robust learning and reinforcement learning. In the paper, we study a class of nonconvex-nonconcave minimax optimization with nonsmooth regularization, where the objective function is possibly nonconvex on primal variable $x$, and it is nonconcave and satisfies the Polyak-Lojasiewicz (PL) condition on dual variable $y$. Moreover, we propose a class of enhanced momentum-based gradient descent ascent methods (i.e., MSGDA and AdaMSGDA) to solve these stochastic nonconvex-PL minimax problems. In particular, our AdaMSGDA algorithm can use various adaptive learning rates in updating the variables $x$ and $y$ without relying on any specifical types. Theoretically, we prove that our methods have the best known sample complexity of $\tilde{O}(ε^{-3})$ only requiring one sample at each loop in finding an $ε$-stationary solution. Some numerical experiments on PL-game and Wasserstein-GAN demonstrate the efficiency of our proposed methods.

math.OC↗

arXiv2025

COSMO: Combination of Selective Memorization for Low-cost Vision-and-Language Navigation

Vision-and-Language Navigation (VLN) tasks have gained prominence within artificial intelligence research due to their potential application in fields like home assistants. Many contemporary VLN approaches, while based on transformer architectures, have increasingly incorporated additional components such as external knowledge bases or map information to enhance performance. These additions, while boosting performance, also lead to larger models and increased computational costs. In this paper, to achieve both high performance and low computational costs, we propose a novel architecture with the COmbination of Selective MemOrization (COSMO). Specifically, COSMO integrates state-space modules and transformer modules, and incorporates two VLN-customized selective state space modules: the Round Selective Scan (RSS) and the Cross-modal Selective State Space Module (CS3). RSS facilitates comprehensive inter-modal interactions within a single scan, while the CS3 module adapts the selective state space module into a dual-stream architecture, thereby enhancing the acquisition of cross-modal interactions. Experimental validations on three mainstream VLN benchmarks, REVERIE, R2R, and R2R-CE, not only demonstrate competitive navigation performance of our model but also show a significant reduction in computational costs.

cs.CV↗

arXiv2025

On The Convergence of Euler Discretization of Finite-Time Convergent Gradient Flows

In this study, we investigate the performance of two novel first-order optimization algorithms, namely the rescaled-gradient flow (RGF) and the signed-gradient flow (SGF). These algorithms are derived from the forward Euler discretization of finite-time convergent flows, comprised of non-Lipschitz dynamical systems, which locally converge to the minima of gradient-dominated functions. We first characterize the closeness between the continuous flows and the discretizations, then we proceed to present (linear) convergence guarantees of the discrete algorithms (in the general and the stochastic case). Furthermore, in cases where problem parameters remain unknown or exhibit non-uniformity, we further integrate the line-search strategy with RGF/SGF and provide convergence analysis in this setting. We then apply the proposed algorithms to academic examples and deep neural network training, our results show that our schemes demonstrate faster convergences against standard optimization alternatives.

cs.LG↗

arXiv2025

FlexVLN: Flexible Adaptation for Diverse Vision-and-Language Navigation Tasks

The aspiration of the Vision-and-Language Navigation (VLN) task has long been to develop an embodied agent with robust adaptability, capable of seamlessly transferring its navigation capabilities across various tasks. Despite remarkable advancements in recent years, most methods necessitate dataset-specific training, thereby lacking the capability to generalize across diverse datasets encompassing distinct types of instructions. Large language models (LLMs) have demonstrated exceptional reasoning and generalization abilities, exhibiting immense potential in robot action planning. In this paper, we propose FlexVLN, an innovative hierarchical approach to VLN that integrates the fundamental navigation ability of a supervised-learning-based Instruction Follower with the robust generalization ability of the LLM Planner, enabling effective generalization across diverse VLN datasets. Moreover, a verification mechanism and a multi-model integration mechanism are proposed to mitigate potential hallucinations by the LLM Planner and enhance execution accuracy of the Instruction Follower. We take REVERIE, SOON, and CVDN-target as out-of-domain datasets for assessing generalization ability. The generalization performance of FlexVLN surpasses that of all the previous methods to a large extent.

cs.CV↗

arXiv2025

Rigidity of infinite inversive distance circle packings in the plane

In 2004, Bowers-Stephenson [2] introduced the inversive distance circle packings as a natural generalization of Thurston's circle packings. They further conjectured the rigidity of infinite inversive distance circle packings in the plane. Motivated by the recent work of Luo-Sun-Wu [22] on Luo's vertex scaling, we prove Bower-Stephenson's conjecture for inversive distance circle packings in the hexagonal triangulated plane. This generalizes Rodin-Sullivan's famous result [13] on the rigidity of infinite tangential circle packings in the hexagonal triangulated plane. The key tools include a maximal principle for generic weighted Delaunay inversive distance circle packings and a ring lemma for the inversive distance circle packings in the hexagonal triangulated plane.

math.GT↗

arXiv2025

Importance-Aware Source-Channel Coding for Multi-Modal Task-Oriented Semantic Communication

This paper explores the concept of information importance in multi-modal task-oriented semantic communication systems, emphasizing the need for high accuracy and efficiency to fulfill task-specific objectives. At the transmitter, generative AI (GenAI) is employed to partition visual data objects into semantic segments, each representing distinct, task-relevant information. These segments are subsequently encoded into tokens, enabling precise and adaptive transmission control. Building on this frame work, we present importance-aware source and channel coding strategies that dynamically adjust to varying levels of significance at the segment, token, and bit levels. The proposed strategies prioritize high fidelity for essential information while permitting controlled distortion for less critical elements, optimizing overall resource utilization. Furthermore, we address the source-channel coding challenge in semantic multiuser systems, particularly in multicast scenarios, where segment importance varies among receivers. To tackle these challenges, we propose solutions such as rate-splitting coded progressive transmission, ensuring flexibility and robustness in task-specific semantic communication.

eess.SP↗

arXiv2024

Constraining the Galactic Structure using Time Domain Gravitational Wave Signal from Double White Dwarfs Detected by Space Gravitational Wave Detectors

The Gravitation Wave (GW) signals from a large number of double white dwarfs (DWDs) in the Galaxy are expected to be detected by space GW detectors, e.g., the Laser Interferometer Space Antenna (LISA), Taiji, and Tianqin in the millihertz band. In this paper, we present an alternative method by directly using the time-domain GW signal detected by space GW detectors to constrain the anisotropic structure of the Galaxy. The information of anisotropic distribution of DWDs is naturally encoded in the time-domain GW signal because of the variation of the detectors' directions and consequently the pattern functions due to their annual motion around the sun. The direct use of the time-domain GW signal enables simple calculations, such as utilizing an analytical method to assess the noise arising from the superposition of random phases of DWDs and using appropriate weights to improve the constraints. We investigate the possible constraints on the scale of the Galactic thin disk and bulge that may be obtained from LISA and Taiji by using this method with mock signals obtained from population synthesis models. We further show the different constraining capabilities of the low-frequency signal (foreground) and the high-frequency signal (resolvable-sources) via the Markov Chain Monte Carlo method, and find that the scale height and length of the Galactic thin disk and the scale radius of bulge can be constrained to a fractional accuracy of ~ 30%, 30%, 40% (or 20%, 10%, 40%) by using the low-frequency (or high-frequency) signal detected by LISA or Taiji.

astro-ph.GA↗