Searcharxiv⌕ Search

arXiv subjects

Xin Liu

Publications and source records attributed to Xin Liu.

At least 163 records · Page 9Linked to original sources
arXiv2026

Study on the detector energy response of SVOM/GRM

The SVOM mission is specifically designed to for the detection and localization of Gamma-Ray Bursts (GRBs) and subsequent follow-up observations. Among the four telescopes installed on the SVOM satellite, the Gamma-Ray Monitor (GRM) plays a crucial role in capturing the prompt emission of GRBs due to its wide field of view (FOV) and broad energy range. Accurate determination of the detector's energy response is vital for analyzing GRM data, particularly considering the significant impact of the atmospheric albedo effect on this response. This research focuses on deriving the detector's energy response and establishing a calibration database for the GRM, with particular emphasis on investigating the atmospheric albedo effect. The study shows that the contribution of albedo photons to the detector's effective area depends strongly on the orientation of the GRD line of sight (LoS) relative to Earth and on the incident direction of the GRB. When the GRD LoS is anti-Earth oriented, the albedo effect is minimal, with the highest proportion of albedo effective area accounting for approximately 10% of the total effective area. This occurs when the incident angle of the GRB is nearly perpendicular to the LoS. Conversely, if the GRD LoS is not pointing away from Earth and the GRB arrives from angles greater than about 90$^{\circ}$, the albedo component can become predominant, contributing up to around 100% of the total effective area. This is especially pronounced in the 8-20 keV range, where the direct effective area drops to zero due to the large GRB injection angle. Our results show that, it is necessary for GRM to consider the atmospheric albedo effects in detector response, otherwise the spectral and localization analyses will result in biased measurements.

astro-ph.HE↗
arXiv2026

The Gamma-Ray Monitor onboard the SVOM satellite

The Gamma-Ray Monitor (GRM) is a key scientific payload onboard the Space-based Multi-band Variable Object Monitor (SVOM) satellite, designed specifically for the detection and study of gamma-ray bursts (GRBs). Launched into a 625 km low-Earth orbit on 22 June 2024, GRM serves as a large-area, wide-field-of-view instrument capable of observing the hard X-ray and soft gamma-ray emissions in the energy range of 15 keV to 5 MeV. Its primary scientific objectives include: promptly triggering and localizing GRBs (with particular sensitivity to short-hard GRBs), measuring spectral and temporal properties of bursts, monitoring charged particle fluxes in orbit. GRM successfully detected its first GRB (GRB 240627B) on 27 June 2024, and has since maintained a detection rate of more than 100 GRBs per year. Cross-instrument comparisons with detectors such as GECAM and Fermi/GBM have validated the performance and data quality of GRM. This paper provides a comprehensive overview of GRM instrument design, reliability verification through ground testing, in-orbit triggering and localization algorithms, performance calibration, and preliminary in-orbit results, demonstrating its capability as a versatile gamma-ray all-sky monitor.

astro-ph.IM↗
arXiv2026

GRM Scientific Pipeline

The Gamma-Ray Monitor (GRM) is a key payload of the Space-based multiband astronomical Variable Objects Monitor (SVOM) mission, which is designed to detect gamma ray bursts (GRBs) within the energy range of 15 keV to 5 MeV. The GRM Instrument Center (GRM\_IC) features real-time data processing through the X-band, enabling rapid response of high-energy GRB events. The system employs an event-driven architecture and distributed design, achieving efficient processing and real-time monitoring of massive observational data. Through comprehensive data production processes and scientific data product management, the system achieves efficient production of scientific data products of the L1B / C level through the submission of jobs to the task scheduling system. Through modular architecture design and automated processing workflow, the GRM data processing system realizes precise conversion and scientific analysis of GRB detection data, providing robust technical support for future system upgrades and cross-platform collaboration.

astro-ph.IM↗
arXiv2026

Coherence toroidal vortices and statistic-veiled correlation topologies

Toroidal vortices in fluid and gas dynamics underpin a broad spectrum of scientific and technological fields, from elementary particle physics to condensed matter systems, and have recently garnered significant attention in optics because of their inherent topological stability. Here we report the experimental observation of toroidal vortices in stochastic optical wavefields with partial coherence, termed coherence toroidal vortices, which eliminates deterministic topological signatures in conventional optical degrees of freedom while unveiling statistically hidden correlation topologies. These underlying topologies-including both fundamental and higher-order hopfionic textures-emerge exclusively in second-order field correlations and are accessible only through statistical measurements. We further examine the impact of chaotic channels on the stability of these statistically veiled correlation topologies, demonstrating that their topological invariants remain robust under realistic environmental perturbations. These findings are experimentally validated and offer novel insights into the potential of toroidal light vortices serving as controllable channels for directional energy and information transfer within complex media.

physics.optics↗
arXiv2026

BackPlay: Head-Only Look-Back Self-Correction for Diffusion Language Models

Diffusion Language Models (DLMs) decode multiple tokens in parallel, but aggressive multi-token decoding amplifies cross-token dependency errors and can sharply degrade generation quality. We propose BackPlay, a frozen-backbone self-correction framework that trains only a lightweight correction head on a finetuned DLM without updating any backbone or adapter parameters. Because the head is trained on errors produced by the same frozen generator used at inference time, its training distribution aligns with the error patterns of the deployed model. We further introduce Look-back Correction, a training mechanism that injects predictions from earlier, more corrupted denoising states into later, richer contexts, enabling the head to leverage later context to detect mistakes made in earlier generation steps. During inference, BackPlay periodically revisits previously generated tokens through selective remasking and regeneration to limit error accumulation. Across mathematical reasoning and code generation benchmarks, BackPlay improves the speed--quality trade-off of the underlying DLM under multi-token decoding.

cs.LG↗
arXiv2026

veScale-FSDP: Flexible and High-Performance FSDP at Scale

Fully Sharded Data Parallel (FSDP), also known as Zero Redundancy Optimizer (ZeRO), is widely used for large-scale model training, because of its memory efficiency and minimal intrusion on model code. However, existing FSDP systems rely on fixed element-wise or row-wise sharding formats that conflict with block-structured computations. As a result, they struggle to support modern structure-aware training methods, including block-wise quantization and non-element-wise optimizers such as Shampoo and Muon. In addition, today's implementations incur communication and memory overheads that degrade efficiency at the scale of tens of thousands of GPUs. We introduce veScale-FSDP, a novel FSDP system that combines RaggedShard, a flexible sharding format, with a structure-aware planning algorithm to deliver both flexibility and performance. veScale-FSDP enables zero-copy FSDP communications and natively supports block-wise quantization and non-element-wise optimizers, achieving 5% to 66% higher throughput and 16% to 30% lower memory usage than existing FSDP systems, while scaling efficiently to tens of thousands of GPUs.

cs.DC↗
arXiv2026

DT2IT-MRM: Debiased Preference Construction and Iterative Training for Multimodal Reward Modeling

Multimodal reward models (MRMs) play a crucial role in aligning Multimodal Large Language Models (MLLMs) with human preferences. Training a good MRM requires high-quality multimodal preference data. However, existing preference datasets face three key challenges: lack of granularity in preference strength, textual style bias, and unreliable preference signals. Besides, existing open-source multimodal preference datasets suffer from substantial noise, yet there is a lack of effective and scalable curation methods to enhance their quality. To address these limitations, we propose \textbf{DT2IT-MRM}, which integrates a \textbf{D}ebiased preference construction pipeline, a novel reformulation of text-to-image (\textbf{T2I}) preference data, and an \textbf{I}terative \textbf{T}raining framework that curates existing multimodal preference datasets for \textbf{M}ultimodal \textbf{R}eward \textbf{M}odeling. Our experimental results show that DT2IT-MRM achieves new \textbf{state-of-the-art} overall performance on three major benchmarks: VL-RewardBench, Multimodal RewardBench, and MM-RLHF-RewardBench.

cs.AI↗
arXiv2026

Logit Arithmetic Elicits Long Reasoning Capabilities Without Training

Large reasoning models exhibit long chain-of-thought reasoning with complex strategies such as backtracking and self-verification. Yet, these capabilities typically require resource-intensive post-training. We investigate whether such behaviors can be elicited in large models without any gradient updates. To this end, we propose a decoding-time approach, ThinkLogit, which utilizes logit arithmetic to transfer these capabilities from a substantially smaller reasoning guider to a large non-reasoning target. We further show that we can boost performance by training the guider to correct the target's errors using preference optimization over mixed model outputs, a setup we refer to as ThinkLogit-DPO. We evaluate these methods across six reasoning benchmarks spanning math, science, and coding domains using the Qwen2.5-32B guided by R1-Distill-Qwen-1.5B, a model 21x smaller. Our experiments demonstrate that ThinkLogit and ThinkLogit-DPO achieve a relative improvement of 21.5% and 24.2%, respectively, over the target model. Moreover, ThinkLogit remains effective even when the guider and target come from different model families. Crucially, our method requires zero training for the large model and would incur minimal inference overhead when logits are computed in parallel, presenting a practical solution for enabling long reasoning at scale.

cs.CL↗
arXiv2026

Momentum Stability and Adaptive Control in Stochastic Reconfiguration

Variational Monte Carlo (VMC) combined with expressive neural network wavefunctions has become a powerful route to high-accuracy ground-state calculations, yet its practical success hinges on efficient and stable wavefunction optimization. While stochastic reconfiguration (SR) provides a geometry-aware preconditioner motivated by imaginary-time evolution, its Kaczmarz-inspired variant, subsampled projected-increment natural gradient descent (SPRING), achieves state-of-the-art empirical performance. However, the effectiveness of SPRING is highly sensitive to the choice of a momentum-like parameter $μ$. The original sensitivity of $μ$ and the instability observed at $μ=1$, have remained unclear. In this work, we clarify the distinct mechanisms governing the regimes $μ<1$ and $μ=1$. We establish convergence guarantees for $0\leμ<1$ under mild assumptions, and construct counterexamples showing that $μ=1$ can induce divergence via uncontrolled growth along kernel-related directions when the step-size is not summable. Motivated by these theoretical insights and numerical observations, we further propose \textit{Principal Range Informed MomEntum SR} (PRIME-SR), a tuning-free momentum-adaptive SR method based on effective spectral dimension and subspace overlap. PRIME-SR achieves performance comparable to optimally tuned SPRING while significantly improving robustness in VMC optimization.

math.OC↗
arXiv2026

Security Risks of VOA-Induced Luminescence in Chip-Based quantum key distribution

Integrated photonics is widely regarded as a key enabler for scalable quantum key distribution (QKD), offering compactness, stability, and compatibility with semiconductor fabrication. Despite rapid advances in chip-based QKD, the implementation security of integrated photonic components remains insufficiently understood. Here we present the first systematic study of an implementation-level security vulnerability associated with p-n junction-based variable optical attenuators (VOAs), a ubiquitous component in integrated QKD transmitters. We theoretically and experimentally demonstrate that electrically biased p-n junction VOAs emit spontaneous luminescence. Using a single-photon-sensitive spectral measurement technique, we identify the emission wavelength to be centered around 1107 nm, well separated from the C-band quantum signals. This spectral separation gives rise to a previously unrecognized wavelength-resolved side channel, enabling potential wavelength-splitting attacks without directly disturbing the encoded quantum states. By incorporating the measured luminescence into a quantitative security analysis, we show that even extremely weak emission can lead to non-negligible information leakage. Our findings reveal a fundamental and previously overlooked security risk in photonic integrated QKD systems and highlight the necessity of security-aware device design for future integrated quantum communication technologies.

quant-ph↗
arXiv2026

FLASH: Fast Learning via GPU-Accelerated Simulation for High-Fidelity Deformable Manipulation in Minutes

Simulation frameworks such as Isaac Sim have enabled scalable robot learning for locomotion and rigid-body manipulation; however, contact-rich simulation remains a major bottleneck for deformable object manipulation. The continuously changing geometry of soft materials, together with large numbers of vertices and contact constraints, makes it difficult to achieve high accuracy, speed, and stability required for large-scale interactive learning. We present FLASH, a GPU-native simulation framework for contact-rich deformable manipulation, built on an accurate NCP-based solver that enforces strict contact and deformation constraints while being explicitly designed for fine-grained GPU parallelism. Rather than porting conventional single-instruction-multiple-data (SIMD) solvers to GPUs, FLASH redesigns the physics engine from the ground up to leverage modern GPU architectures, including optimized collision handling and memory layouts. As a result, FLASH scales to over 3 million degrees of freedom at 30 FPS on a single RTX 5090, while accurately simulating physical interactions. Policies trained solely on FLASH-generated synthetic data in minutes achieve robust zero-shot sim-to-real transfer, which we validate on physical robots performing challenging deformable manipulation tasks such as towel folding and garment folding, without any real-world demonstration, providing a practical alternative to labor-intensive real-world data collection.

cs.RO↗
arXiv2026

Detecting LLM-Generated Spam Reviews by Integrating Language Model Embeddings and Graph Neural Network

The rise of large language models (LLMs) has enabled the generation of highly persuasive spam reviews that closely mimic human writing. These reviews pose significant challenges for existing detection systems and threaten the credibility of online platforms. In this work, we first create three realistic LLM-generated spam review datasets using three distinct LLMs, each guided by product metadata and genuine reference reviews. Evaluations by GPT-4.1 confirm the high persuasion and deceptive potential of these reviews. To address this threat, we propose FraudSquad, a hybrid detection model that integrates text embeddings from a pre-trained language model with a gated graph transformer for spam node classification. FraudSquad captures both semantic and behavioral signals without relying on manual feature engineering or massive training resources. Experiments show that FraudSquad outperforms state-of-the-art baselines by up to 44.22% in precision and 43.01% in recall on three LLM-generated datasets, while also achieving promising results on two human-written spam datasets. Furthermore, FraudSquad maintains a modest model size and requires minimal labeled training data, making it a practical solution for real-world applications. Our contributions include new synthetic datasets, a practical detection framework, and empirical evidence highlighting the urgency of adapting spam detection to the LLM era. Our code and datasets are available at: https://anonymous.4open.science/r/FraudSquad-5389/.

cs.CL↗
arXiv2026

Toward Exact Convergence in Byzantine-Robust Decentralized Learning: A Statistical Identification Approach

To defend against Byzantine attacks in decentralized learning, most existing methods rely on robust aggregation rules to mitigate the influence of malicious machines. However, these strategies inherently introduce bias, leading to inexact convergence with non-vanishing steady-state errors. In this paper, we propose a strategic shift from passive aggregation to active identification by introducing the Decentralized Rescaled Stochastic Gradient Descent with Byzantine Machine Identification (DRSGD-ByMI) framework. The core of our approach is an identification-based ``detect-then-optimize'' pipeline, where a p-value-free detection procedure is developed to accurately prune malicious nodes from the network. By leveraging sample-splitting score statistics, this identification mechanism achieves false discovery rate control without requiring restrictive distributional assumptions. We theoretically demonstrate that this precise identification allows the decentralized network to recover sufficient connectivity among the normal nodes, thereby enabling DRSGD-ByMI to match, even in the presence of Byzantine machines, the same order-optimal convergence rate as standard decentralized stochastic first-order methods. Numerical experiments validate our theoretical results and demonstrate the effectiveness of DRSGD-ByMI for decentralized robust learning problems.

stat.ME↗
arXiv2026

Agentic Conversational Search with Contextualized Reasoning via Reinforcement Learning

Large Language Models (LLMs) have become a popular interface for human-AI interaction, supporting information seeking and task assistance through natural, multi-turn dialogue. To respond to users within multi-turn dialogues, the context-dependent user intent evolves across interactions, requiring contextual interpretation, query reformulation, and dynamic coordination between retrieval and generation. Existing studies usually follow static rewrite, retrieve, and generate pipelines, which optimize different procedures separately and overlook the mixed-initiative action optimization simultaneously. Although the recent developments in deep search agents demonstrate the effectiveness in jointly optimizing retrieval and generation via reasoning, these approaches focus on single-turn scenarios, which might lack the ability to handle multi-turn interactions. We introduce a conversational agent that interleaves search and reasoning across turns, enabling exploratory and adaptive behaviors learned through reinforcement learning (RL) training with tailored rewards towards evolving user goals. The experimental results across four widely used conversational benchmarks demonstrate the effectiveness of our methods by surpassing several existing strong baselines.

cs.CL↗
arXiv2026

Scalable Trajectory Generation for Whole-Body Mobile Manipulation

Robots deployed in unstructured environments must coordinate whole-body motion -- simultaneously moving a mobile base and arm -- to interact with the physical world. This coupled mobility and dexterity yields a state space that grows combinatorially with scene and object diversity, demanding datasets far larger than those sufficient for fixed-base manipulation. Yet existing acquisition methods, including teleoperation and planning, are either labor-intensive or computationally prohibitive at scale. The core bottleneck is the lack of a scalable pipeline for generating large-scale, physically valid, coordinated trajectory data across diverse embodiments and environments. Here we introduce AutoMoMa, a GPU-accelerated framework that unifies AKR modeling, which consolidates base, arm, and object kinematics into a single chain, with parallelized trajectory optimization. AutoMoMa achieves 5,000 episodes per GPU-hour (over $80\times$ faster than CPU-based baselines), producing a dataset of over 500k physically valid trajectories spanning 330 scenes, diverse articulated objects, and multiple robot embodiments. Prior datasets were forced to compromise on scale, diversity, or kinematic fidelity; AutoMoMa addresses all three simultaneously. Training downstream IL policies further reveals that even a single articulated-object task requires tens of thousands of demonstrations for SOTA methods to reach $\approx 80\%$ success, confirming that data scarcity -- not algorithmic limitations -- has been the binding constraint. AutoMoMa thus bridges high-performance planning and reliable IL-based control, providing the infrastructure previously missing for coordinated mobile manipulation research. By making large-scale, kinematically valid training data practical, AutoMoMa showcases generalizable whole-body robot policies capable of operating in the diverse, unstructured settings of the real world.

cs.RO↗
arXiv2026

DeepDISC-Euclid: Source Classification and Photometric Redshifts in Euclid Deep Field North With a Pixel-Level Deep Learning Approach

The first Euclid Quick Data Release (Q1) provides extensive imaging and spectroscopic data for hundreds of millions of photometric objects across several deep fields. Accurate classifications and photometric redshifts (photo-z) for these sources are crucial to maximizing the value of these data. In this work, we perform source classification and photo-z estimation for the Euclid Deep Field North (EDF-N) around the North Ecliptic Pole, using a deep learning framework (DeepDISC) that learns and infers using 9-band images simultaneously. We train three dedicated models for (1) source detection and classification, (2) galaxy photo-z, and (3) quasar photo-z. The Euclid Q1 input source catalog, and classifications and spectroscopic redshifts (spec-z) from the Dark Energy Spectroscopic Instrument Data Release 1 are adopted as our training data. DeepDISC source detection achieves overall completeness of ~93% and purity of ~80% if using the Euclid source catalog as the ground truth. Using a JWST source catalog within EDF-N as the reference, we estimate a true purity of ~ 90% for DeepDISC sources. About 99.2%, 99.0%, and 84.8% of stars, galaxies, and quasars, respectively, are correctly recovered with their spectroscopic classifications. The DeepDISC photo-zs show good agreement with spectroscopic redshifts, for both galaxies and quasars. Comparisons with other Euclid Q1 products demonstrate that DeepDISC provides comparable or improved performance in source detection/deblending, classification and photo-z, especially for quasars. These results demonstrate the potential of pixel-level deep learning approaches for large-scale sky surveys such as Euclid and Roman, which will continue to improve with better training labels. We release the full DeepDISC source catalog (~13 million objects) for EDF-N with classifications and photo-zs, including photo-z probability distributions.

astro-ph.GA↗
arXiv2026

Think Through Uncertainty: Improving Long-Form Generation Factuality via Reasoning Calibration

Large language models (LLMs) often hallucinate in long-form generation. Existing approaches mainly improve factuality through post-hoc revision or reinforcement learning (RL) with correctness-based rewards, but they do not teach the model to estimate which parts of its generation are reliable. As a result, models may still state incorrect claims confidently in their responses. Recent advances in reasoning have significantly improved LLM performance, and have been leveraged to estimate confidence by incorporating calibration into RL objectives. However, existing approaches remain limited to a single scalar confidence for the entire response, which is insufficient for long-form generation where uncertainty varies across individual claims. To mitigate this problem, we propose CURE, a framework that improves long-form factuality by teaching LLMs to reason about uncertainty at the claim level. We first introduce a Claim-Aware Reasoning Protocol, which structures outputs into atomic claims paired with explicit confidence estimates. We then develop a multi-stage training pipeline that aligns model confidence with claims' correctness and then optimizes on factuality. The resulting calibrated confidence further enables selective prediction, allowing the model to abstain from uncertain claims at inference time. Experiments on four long-form factuality benchmarks show that CURE consistently improves factual accuracy over competitive supervised and RL baselines, while maintaining factual recall. In particular, it improves claim-level accuracy by up to 39.9% on Biography generation. These gains are accompanied by improved calibration, as reflected by a 16.0% increase in AUROC on FactBench.

cs.CL↗
arXiv2026

SessionIntentBench: A Multi-task Inter-session Intention-shift Modeling Benchmark for E-commerce Customer Behavior Understanding

Session history is a common way of recording user interacting behaviors throughout a browsing activity with multiple products. For example, if an user clicks a product webpage and then leaves, it might because there are certain features that don't satisfy the user, which serve as an important indicator of on-the-spot user preferences. However, all prior works fail to capture and model customer intention effectively because insufficient information exploitation and only apparent information like descriptions and titles are used. There is also a lack of data and corresponding benchmark for explicitly modeling intention in E-commerce product purchase sessions. To address these issues, we introduce the concept of an intention tree and propose a dataset curation pipeline. Together, we construct a sibling multimodal benchmark, SessionIntentBench, that evaluates L(V)LMs' capability on understanding inter-session intention shift with four subtasks. With 1,952,177 intention entries, 1,132,145 session intention trajectories, and 13,003,664 available tasks mined using 10,905 sessions, we provide a scalable way to exploit the existing session data for customer intention understanding. We conduct human annotations to collect ground-truth label for a subset of collected data to form an evaluation gold set. Extensive experiments on the annotated data further confirm that current L(V)LMs fail to capture and utilize the intention across the complex session setting. Further analysis show injecting intention enhances LLMs' performances.

cs.CL↗