arXiv subjects

Xiang Liu

Publications and source records attributed to Xiang Liu.

At least 127 records · Page 7Linked to original sources

arXiv2025

Reasoning Language Model Inference Serving Unveiled: An Empirical Study

The reasoning large language model (RLLM) has been proven competitive in solving complex reasoning tasks such as mathematics, coding, compared to general LLM. However, the serving performance and behavior of RLLM remains unexplored, which may undermine the deployment and utilization of RLLM in real-world scenario. To close this gap, in this paper, we conduct a comprehensive study of RLLM service. We first perform a pilot study on comparing the serving performance between RLLM and traditional LLM and reveal that there are several distinct differences regarding serving behavior: (1) significant memory usage and fluctuations; (2) straggler requests; (3) adaptive running time; (4) domain preference. Then we further investigate whether existing inference optimization techniques are valid for RLLM. Our main takeaways are that model quantization methods and speculative decoding can improve service system efficiency with small compromise to RLLM accuracy, while prefix caching, KV cache quantization may even degrade accuracy or serving performance for small RLLM. Lastly, we conduct evaluation under real world workload modeled by Gamma distribution to verify our findings. Empirical results of real world workload evaluation across different dataset are aligned with our main findings regarding RLLM serving. We hope our work can provide the research community and industry with insights to advance RLLM inference serving.

cs.LG↗

arXiv2025

Styl3R: Instant 3D Stylized Reconstruction for Arbitrary Scenes and Styles

Stylizing 3D scenes instantly while maintaining multi-view consistency and faithfully resembling a style image remains a significant challenge. Current state-of-the-art 3D stylization methods typically involve computationally intensive test-time optimization to transfer artistic features into a pretrained 3D representation, often requiring dense posed input images. In contrast, leveraging recent advances in feed-forward reconstruction models, we demonstrate a novel approach to achieve direct 3D stylization in less than a second using unposed sparse-view scene images and an arbitrary style image. To address the inherent decoupling between reconstruction and stylization, we introduce a branched architecture that separates structure modeling and appearance shading, effectively preventing stylistic transfer from distorting the underlying 3D scene structure. Furthermore, we adapt an identity loss to facilitate pre-training our stylization model through the novel view synthesis task. This strategy also allows our model to retain its original reconstruction capabilities while being fine-tuned for stylization. Comprehensive evaluations, using both in-domain and out-of-domain datasets, demonstrate that our approach produces high-quality stylized 3D content that achieve a superior blend of style and scene appearance, while also outperforming existing methods in terms of multi-view consistency and efficiency.

cs.CV↗

arXiv2025

AnTKV: Anchor Token-Aware Sub-Bit Vector Quantization for KV Cache in Large Language Models

Quantization has emerged as an effective and lightweight solution to reduce the memory footprint of the KV cache in Large Language Models. Nevertheless, minimizing the accuracy degradation caused by ultra-low-bit KV cache quantization remains a significant challenge. While scalar quantization is constrained by 1-bit bound, vector quantization exploits intra-vector correlations and enables sub-bit regimes, making it more suitable for ultra-low-bit quantization. To further mitigate quantization-induced degradation, we reveal that the degradation is highly uneven across tokens in attention quality. To investigate this unevenness, we introduce anchor score to measure each token's sensitivity to quantization. Our analysis and experiments show that preserving a small subset (1\%) of tokens with the highest Anchor Score significantly mitigates accuracy loss under aggressive quantization. We propose AnTKV, a dual-stage framework that leverages anchor token-aware vector quantization to compress the KV cache. It combines offline token-aware centroids learning and online anchor token selection to balance compression and accuracy. To enable efficient deployment, we design an online anchor token selection kernel compatible with FlashAttention. It allows LLaMA3-8B to scale to 840K tokens on a single 80GB A100, while delivering up to $3.5\times$ higher decoding throughput over the FP16 baseline. Experiments demonstrate that AnTKV matches or surpasses prior methods at 4-bit, and significantly reduce perplexity under ultra-low-bit quantization, achieving 6.32 at 1-bit on Mistral-7B, compared to 7.25 for CQ and 15.36 for KVQuant.

cs.CL↗

arXiv2025

ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference

Large Language Models (LLMs) require significant GPU memory when processing long texts, with the key value (KV) cache consuming up to 70\% of total memory during inference. Although existing compression methods reduce memory by evaluating the importance of individual tokens, they overlook critical semantic relationships between tokens, resulting in fragmented context and degraded performance. We introduce ChunkKV, which fundamentally reimagines KV cache compression by treating semantic chunks - rather than isolated tokens - as basic compression units. This approach preserves complete linguistic structures and contextual integrity, ensuring that essential meaning is retained even under aggressive compression. Our innovation includes a novel layer-wise index reuse technique that exploits the higher cross-layer similarity of preserved indices in ChunkKV, reducing computational overhead and improving throughput by 26.5\%. Comprehensive evaluations on challenging benchmarks: LongBench, Needle-In-A-HayStack, GSM8K, and JailbreakV demonstrate that ChunkKV outperforms state-of-the-art methods by up to 8.7\% in precision while maintaining the same compression ratio. These results confirm that semantic-aware compression significantly enhances both efficiency and performance for long-context LLM inference, providing a simple yet effective solution to the memory bottleneck problem. The code is available at \href{https://github.com/NVIDIA/kvpress}{link}.

cs.CL↗

arXiv2025

Medium modifications of $1P$-wave charmonia $χ_{cJ}(1P)$ in cold nuclear matter

In this work, we employ the quark-meson coupling model to investigate the mass shifts of $1P$-wave charmonia $χ_{cJ}(1P)$ ($J=0,1,2$) in cold symmetric nuclear matter by incorporating in-medium loop contributions to the $χ_{cJ}(1P)$ self-energy within the unquenched picture. At normal nuclear matter density, we obtain significant mass reductions of about 60 MeV for the $χ_{cJ}(1P)$ states, with the $χ_{c2}(1P)$ mass shift primarily arising from the vector-vector loop. Our results also indicate the absence of level crossing between the in-medium $χ_{cJ}(1P)$ mass and the $D\bar{D}$ mass threshold up to $ρ_B < 3ρ_0$-a feature that could be probed in the Compressed Baryonic Matter experiment at FAIR and the Beam Energy Scan program at RHIC.

hep-ph↗

arXiv2025

Perovskite-LLM: Knowledge-Enhanced Large Language Models for Perovskite Solar Cell Research

The rapid advancement of perovskite solar cells (PSCs) has led to an exponential growth in research publications, creating an urgent need for efficient knowledge management and reasoning systems in this domain. We present a comprehensive knowledge-enhanced system for PSCs that integrates three key components. First, we develop Perovskite-KG, a domain-specific knowledge graph constructed from 1,517 research papers, containing 23,789 entities and 22,272 relationships. Second, we create two complementary datasets: Perovskite-Chat, comprising 55,101 high-quality question-answer pairs generated through a novel multi-agent framework, and Perovskite-Reasoning, containing 2,217 carefully curated materials science problems. Third, we introduce two specialized large language models: Perovskite-Chat-LLM for domain-specific knowledge assistance and Perovskite-Reasoning-LLM for scientific reasoning tasks. Experimental results demonstrate that our system significantly outperforms existing models in both domain-specific knowledge retrieval and scientific reasoning tasks, providing researchers with effective tools for literature review, experimental design, and complex problem-solving in PSC research.

cs.AI↗

arXiv2025

FlowKV: Enhancing Multi-Turn Conversational Coherence in LLMs via Isolated Key-Value Cache Management

Large Language Models (LLMs) are increasingly deployed in multi-turn conversational applications, where the management of the Key-Value (KV) Cache presents a significant bottleneck. The linear growth of the KV Cache with dialogue history imposes substantial computational costs, and existing eviction strategies often degrade performance by repeatedly compressing early conversational context, leading to information loss and context forgetting. This paper introduces FlowKV, a novel \textbf{multi-turn isolation mechanism} for KV Cache management, which can be applied to any KV Cache compression method without training. FlowKV's core innovation is a multi-turn isolation mechanism that preserves the accumulated compressed KV cache from past turns. Compression is then strategically applied only to the newly generated KV pairs of the latest completed turn, effectively preventing the re-compression of older context and thereby mitigating catastrophic forgetting. Our results demonstrate that FlowKV consistently and significantly outperforms baseline strategies in maintaining instruction-following accuracy and user preference retention from 10.90\% to 75.40\%, particularly in later conversational turns.

cs.CL↗

arXiv2025

Triply heavy baryon spectroscopy revisited

We present a comprehensive study of triply heavy baryons ($Ω_{ccc}$, $Ω_{bbb}$, $Ω_{bcc}$, and $Ω_{bbc}$) within the nonrelativistic quark model, employing the Gaussian expansion method to calculate mass spectra up to $D$-wave states. Our analysis represents the most complete treatment to date for this model, incorporating full angular momentum mixing effects. While our predictions for low-lying states agree well with lattice quantum chromodynamics (QCD) results, we find systematically lower masses for excited states compared to lattice calculations. Using the obtained wave functions, we estimate radiative decay widths up to $1D$ states, revealing significant differences from previous theoretical work. Additionally, we identify and resolve several misconceptions in prior treatments of triply heavy baryon spectroscopy, particularly symmetry constraint and wave function construction in three-quark systems. These results provide crucial information for future experimental searches and theoretical investigations of triply heavy baryon systems.

hep-ph↗

arXiv2025

Can Data-Driven Dynamics Reveal Hidden Physics? There Is A Need for Interpretable Neural Operators

Recently, neural operators have emerged as powerful tools for learning mappings between function spaces, enabling data-driven simulations of complex dynamics. Despite their successes, a deeper understanding of their learning mechanisms remains underexplored. In this work, we classify neural operators into two types: (1) Spatial domain models that learn on grids and (2) Functional domain models that learn with function bases. We present several viewpoints based on this classification and focus on learning data-driven dynamics adhering to physical principles. Specifically, we provide a way to explain the prediction-making process of neural operators and show that neural operator can learn hidden physical patterns from data. However, this explanation method is limited to specific situations, highlighting the urgent need for generalizable explanation methods. Next, we show that a simple dual-space multi-scale model can achieve SOTA performance and we believe that dual-space multi-spatio-scale models hold significant potential to learn complex physics and require further investigation. Lastly, we discuss the critical need for principled frameworks to incorporate known physics into neural operators, enabling better generalization and uncovering more hidden physical phenomena.

cs.LG↗

arXiv2025

Fuzzy Reasoning Chain (FRC): An Innovative Reasoning Framework from Fuzziness to Clarity

With the rapid advancement of large language models (LLMs), natural language processing (NLP) has achieved remarkable progress. Nonetheless, significant challenges remain in handling texts with ambiguity, polysemy, or uncertainty. We introduce the Fuzzy Reasoning Chain (FRC) framework, which integrates LLM semantic priors with continuous fuzzy membership degrees, creating an explicit interaction between probability-based reasoning and fuzzy membership reasoning. This transition allows ambiguous inputs to be gradually transformed into clear and interpretable decisions while capturing conflicting or uncertain signals that traditional probability-based methods cannot. We validate FRC on sentiment analysis tasks, where both theoretical analysis and empirical results show that it ensures stable reasoning and facilitates knowledge transfer across different model scales. These findings indicate that FRC provides a general mechanism for managing subtle and ambiguous expressions with improved interpretability and robustness.

cs.CL↗

arXiv2025

Doubly charmed hexaquarks in the diquark picture

We investigate doubly charmed hexaquark states within the diquark picture, by employing the constituent quark model and the quark-interchange model as our theoretical frameworks. Using the Gaussian expansion method, we systematically study these states, with calculating various properties such as mass spectra, internal contributions of each Hamiltonian component, root-mean-square radii, and two-body strong decay widths. Our analysis of the mass spectra reveals no stable state in this system. Furthermore, the root-mean-square radii suggest that the doubly charmed hexaquark states exhibit a compact configuration. By examining the decay widths, we identify potentially detectable states and their primary decay channels within each subsystem. Despite the large decay phase space, we still find narrow states with total widths of less than 10 MeV. This study provides a theoretical foundation for understanding the structures and interactions of doubly charmed hexaquark states and offers valuable insights for future experimental searches.

hep-ph↗

arXiv2025

Artificial ferroelectric-like hysteresis in antiferroelectrics with non-uniform disorder

Antiferroelectrics exhibit unique double-hysteresis polarization loops, which have garnered significant attention due to their potential applications such as energy storage, electromechanical transduction, as well as synapse devices. However, numerous antiferroelectric materials have been reported to display signs of hysteresis loops resembling those of ferroelectric materials, and a comprehensive understanding remains elusive. In this work, we provide a phenomenological model that reproduces such widely observed artificial ferroelectric hysteresis with a superposition of numerous disordered antiferroelectric loops that have varying antiferroelectric-to-ferroelectric transition fields, particularly when these field ranges intersect. Experimentally, we realized such artificial ferroelectric-like hysteresis loops in the prototypical antiferroelectric PbZrO$_3$ and PbHfO$_3$ thin films, by introducing non-uniform local disorder (e.g., defects) via fine-tuning of the film growth conditions. These ferroelectric-like states are capable of persisting for several hours prior to transitioning back into the thermodynamically stable antiferroelectric ground state. Those results provide insights into the fundamental impact of disorder on the AFE properties and new possibilities of disorder-tailored functions.

cond-mat.mtrl-sci↗

arXiv2025

Hierarchical Deep Fusion Framework for Multi-dimensional Facial Forgery Detection -- The 2024 Global Deepfake Image Detection Challenge

The proliferation of sophisticated deepfake technology poses significant challenges to digital security and authenticity. Detecting these forgeries, especially across a wide spectrum of manipulation techniques, requires robust and generalized models. This paper introduces the Hierarchical Deep Fusion Framework (HDFF), an ensemble-based deep learning architecture designed for high-performance facial forgery detection. Our framework integrates four diverse pre-trained sub-models, Swin-MLP, CoAtNet, EfficientNetV2, and DaViT, which are meticulously fine-tuned through a multi-stage process on the MultiFFDI dataset. By concatenating the feature representations from these specialized models and training a final classifier layer, HDFF effectively leverages their collective strengths. This approach achieved a final score of 0.96852 on the competition's private leaderboard, securing the 20th position out of 184 teams, demonstrating the efficacy of hierarchical fusion for complex image classification tasks.

cs.CV↗

arXiv2025

End to End Autoencoder MLP Framework for Sepsis Prediction

Sepsis is a life threatening condition that requires timely detection in intensive care settings. Traditional machine learning approaches, including Naive Bayes, Support Vector Machine (SVM), Random Forest, and XGBoost, often rely on manual feature engineering and struggle with irregular, incomplete time-series data commonly present in electronic health records. We introduce an end-to-end deep learning framework integrating an unsupervised autoencoder for automatic feature extraction with a multilayer perceptron classifier for binary sepsis risk prediction. To enhance clinical applicability, we implement a customized down sampling strategy that extracts high information density segments during training and a non-overlapping dynamic sliding window mechanism for real-time inference. Preprocessed time series data are represented as fixed dimension vectors with explicit missingness indicators, mitigating bias and noise. We validate our approach on three ICU cohorts. Our end-to-end model achieves accuracies of 74.6 percent, 80.6 percent, and 93.5 percent, respectively, consistently outperforming traditional machine learning baselines. These results demonstrate the framework's superior robustness, generalizability, and clinical utility for early sepsis detection across heterogeneous ICU environments.

cs.LG↗

arXiv2025

$D^{()}\bar{B}^{()}$ dynamics in chiral effective field theory

In this work, we systematically study the interactions of the $S$-wave $D^{(*)}\bar{B}^{(*)}$ systems within the framework of chiral effective field theory in heavy hadron formalism. We calculate the $D^{(*)}\bar{B}^{(*)}$ effective potentials up to next-to-leading order, explore the bound state formations, and investigate the $D^{(*)}\bar{B}^{(*)}$ scattering properties such as scattering rate, scattering length, and effective range. Our results show that all $I=1$ $D^{(*)}\bar{B}^{(*)}$ potentials are repulsive, preventing the formation of bound states, while the $I=0$ potentials are generally attractive. Specifically, we get two important observations: first, the shallow bound state is more likely to exist in the $D\bar{B}[I(J^{P})=0(0^{+})]$ system than in the $D\bar{B}^{*}[I(J^{P})=0(1^{+})]$ system; second, $D^{*}\bar{B}^{*}[I(J^{P})=0(0^{+})]$ and $D^{*}\bar{B}^{*}[I(J^{P})=0(1^{+})]$ systems possess relatively large binding energies and positive scattering lengths, which suggests strong bound state formations in these channels. So the attractions in the $D^{*}\bar{B}^{*}[I=0]$ systems are deeper than those in the $D\bar{B}^{(*)}[I=0]$ systems, thus we strongly recommend the future experiment to search for the $D^{*}\bar{B}^{*}[I=0]$ tetraquark systems. We also investigate the coupled-channel effects on the $J=0, 1$ systems and conclude that the inclusion of the coupled channels introduces small but visible influences. In addition, we also investigate the dependencies of the $D\bar{B}^{(*)}$ binding energies on the contact low-energy coupling constants.

hep-ph↗

arXiv2025

Topological Learning Prediction of Virus-like Particle Stoichiometry and Stability

Understanding the stoichiometry and associated stability of virus-like particles (VLPs) is crucial for optimizing their assembly efficiency and immunogenic properties, which are essential for advancing biotechnology, vaccine design, and drug delivery. However, current experimental methods for determining VLP stoichiometry are labor-intensive, and time consuming. Machine learning approaches have hardly been applied to the study of VLPs. To address this challenge, we introduce a novel persistent Laplacian-based machine learning (PLML) mode that leverages both harmonic and non-harmonic spectra to capture intricate topological and geometric features of VLP structures. This approach achieves superior performance on the VLP200 dataset compared to existing methods. To further assess robustness and generalizability, we collected a new dataset, VLP706, containing 706 VLP samples with expanded stoichiometry diversity. Our PLML model maintains strong predictive accuracy on VLP706. Additionally, through random sequence perturbative mutation analysis, we found that 60-mers and 180-mers exhibit greater stability than 240-mers and 420-mers.

q-bio.BM↗

arXiv2025

Identifying triple-strangeness $Ω$ hyperons in light of recent experimental results

In this work, we systematically calculate the mass spectra and decay widths of $Ω$ baryons with $N\leq3$, where $N=2n_ρ+2n_λ+l_ρ+l_λ$ is the shell of the three-body harmonic oscillator. In our scheme, the newly observed $Ω(2109)$ is identified as a $N=2$ candidate, with its quantum numbers tentatively assigned as $Ω(2S,3/2^+)$. However, the experimentally measured mass exhibits a deviation of approximately 50 MeV below the theoretical predictions. The possible reason for this deviation, along with the suggestion for experiments, are discussed in this paper. Additionally, we give possible assignments of $Ω(2012)^-$, $Ω(2250)^-$, $Ω(2380)^-$, and $Ω(2470)^-$. Furthermore, we calculate the masses and widths of other missing $Ω$ states, which could provide valuable clues for experimental searches aimed at discovering these states.

hep-ph↗

arXiv2025

Implication of neutron star observations to the origin of nucleon mass

We investigate the implications of neutron star observations for understanding the origin of nucleon mass using a framework that combines three complementary approaches: the equation of state based on parity doublet structure for hadronic matter below $2n_0$, the Nambu-Jona-Lasinio (NJL) model for quark matter above $5n_0$, and a model-independent analysis of the intermediate density region based on fundamental physical principles. By systematically exploring parameter spaces and comparing theoretical predictions with recent observational constraints, we establish constraints on the chiral invariant mass. Our results suggest that more than a half of the nucleon mass originates from sources beyond spontaneous chiral symmetry breaking, challenging conventional understanding of nucleon mass generation. These constraints arise solely from fundamental physical principles and observational data, independent of specific assumptions about the nature of the quark-hadron transition, providing robust insights into the microscopic origin of hadron masses.

nucl-th↗