arXiv subjects

Hong Liu

Publications and source records attributed to Hong Liu.

At least 109 records · Page 6Linked to original sources

arXiv2025

LongCat-Flash Technical Report

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

cs.CL↗

arXiv2025

Multi-modal Adaptive Estimation for Temporal Respiratory Disease Outbreak

Timely and robust influenza incidence forecasting is critical for public health decision-making. This paper presents MAESTRO (Multi-modal Adaptive Estimation for Temporal Respiratory Disease Outbreak), a novel, unified framework that synergistically integrates advanced spectro-temporal modeling with multi-modal data fusion, including surveillance, web search trends, and meteorological data. By adaptively weighting heterogeneous data sources and decomposing complex time series patterns, the model achieves robust and accurate forecasts. Evaluated on over 11 years of Hong Kong influenza data (excluding the COVID-19 period), MAESTRO demonstrates state-of-the-art performance, achieving a superior model fit with an R-square of 0.956. Extensive ablations confirm the significant contributions of its multi-modal and spectro-temporal components. The modular and reproducible pipeline is made publicly available to facilitate deployment and extension to other regions and pathogens, presenting a powerful tool for epidemiological forecasting.

cs.LG↗

arXiv2025

Enhancing WSI-Based Survival Analysis with Report-Auxiliary Self-Distillation

Survival analysis based on Whole Slide Images (WSIs) is crucial for evaluating cancer prognosis, as they offer detailed microscopic information essential for predicting patient outcomes. However, traditional WSI-based survival analysis usually faces noisy features and limited data accessibility, hindering their ability to capture critical prognostic features effectively. Although pathology reports provide rich patient-specific information that could assist analysis, their potential to enhance WSI-based survival analysis remains largely unexplored. To this end, this paper proposes a novel Report-auxiliary self-distillation (Rasa) framework for WSI-based survival analysis. First, advanced large language models (LLMs) are utilized to extract fine-grained, WSI-relevant textual descriptions from original noisy pathology reports via a carefully designed task prompt. Next, a self-distillation-based pipeline is designed to filter out irrelevant or redundant WSI features for the student model under the guidance of the teacher model's textual knowledge. Finally, a risk-aware mix-up strategy is incorporated during the training of the student model to enhance both the quantity and diversity of the training data. Extensive experiments carried out on our collected data (CRC) and public data (TCGA-BRCA) demonstrate the superior effectiveness of Rasa against state-of-the-art methods. Our code is available at https://github.com/zhengwang9/Rasa.

cs.CV↗

arXiv2025

Ensembling Large Language Models for Code Vulnerability Detection: An Empirical Evaluation

Code vulnerability detection is crucial for ensuring the security and reliability of modern software systems. Recently, Large Language Models (LLMs) have shown promising capabilities in this domain. However, notable discrepancies in detection results often arise when analyzing identical code segments across different training stages of the same model or among architecturally distinct LLMs. While such inconsistencies may compromise detection stability, they also highlight a key opportunity: the latent complementarity among models can be harnessed through ensemble learning to create more robust vulnerability detection systems. In this study, we explore the potential of ensemble learning to enhance the performance of LLMs in source code vulnerability detection. We conduct comprehensive experiments involving five LLMs (i.e., DeepSeek-Coder-6.7B, CodeLlama-7B, CodeLlama-13B, CodeQwen1.5-7B, and StarCoder2-15B), using three ensemble strategies (i.e., Bagging, Boosting, and Stacking). These experiments are carried out across three widely adopted datasets (i.e., Devign, ReVeal, and BigVul). Inspired by Mixture of Experts (MoE) techniques, we further propose Dynamic Gated Stacking (DGS), a Stacking variant tailored for vulnerability detection. Our results demonstrate that ensemble approaches can significantly improve detection performance, with Boosting excelling in scenarios involving imbalanced datasets. Moreover, DGS consistently outperforms traditional Stacking, particularly in handling class imbalance and multi-class classification tasks. These findings offer valuable insights into building more reliable and effective LLM-based vulnerability detection systems through ensemble learning.

cs.SE↗

arXiv2025

On the algebraic stretching dynamics of variable-density mixing in shock-bubble interaction

The mixing mechanism within a single-vortex has been a theoretical focus for decades, while remains unclear especially under variable-density (VD) scenario. This study investigates canonical single-vortex VD mixing in shock-bubble interactions (SBI) through high-resolution numerical simulations. Special attention is paid to examine the stretching dynamics and its impact on VD mixing within a single-vortex, and this problem is investigated by quantitatively characterizing the scalar dissipation rate (SDR), namely mixing rate, and its time integral, referred to as mixedness. We first examine single-vortex passive-scalar (PS) mixing with the absence of density difference. Under several assumptions, the single-vortex stretching rate illustrates an algebraic growth of the length of scalar strips. By incorporating diffusion process through the solution of the advection-diffusion equation along these stretched strips, a PS mixing model for SDR is proposed. Within this framework, density-gradient effects from two perspectives of stretching dynamics and diffusion process are discovered to challenge the extension of PS mixing model to VD mixing. First, the secondary baroclinic effect increases the VD stretching rate by the additional secondary baroclinic principal strain. Second, the density source effect suppresses the diffusion process. By accounting for both the secondary baroclinic effect on stretching and the density source effect on diffusion, a VD mixing model for SBI is further modified. This model establishes a quantitative relationship between the stretching dynamics and the evolution of the mixing rate and mixedness for single-vortex VD mixing over a broad range of Mach numbers. Furthermore, the essential role of stretching dynamics on the mixing rate is demonstrated by the derived dependence of the time-averaged mixing rate on the Peclet number Pe.

physics.flu-dyn↗

arXiv2025

Towards a holographic description of closed universes

We study baby closed universes in AdS, focusing on the Antonini-Sasieta-Swingle (AS$^2$) cosmology, which arises in the gravitational description of partially entangled thermal states (PETS), as well as the classical example of Maldacena-Maoz (MM). We show that the algebraic formulation of AdS/CFT--and in particular the subregion-subalgebra duality--provides a natural framework for describing such universes within the standard AdS/CFT setting, phrased in terms of an operator algebra ${\cal M}$ and a state $ω$ on that algebra, with no need to introduce observers. The algebra encodes all physical operations in the closed universe, and, in principle, all physical observables are obtainable from the pair $({\cal M},ω)$. Along the way, we propose a resolution to a puzzle raised by Antonini and Rath (AR) and reinforced by recent arguments of Engelhardt and Gesteau and a no-go theorem of Gesteau: that a semiclassical baby universe in the AS$^2$ cosmology cannot be understood from the boundary in the usual manner. Our analysis motivates an averaged large-$N$ limit as part of the AdS/CFT dictionary and points toward a unified treatment of spacetimes of all asymptotics in terms of operator algebras. Finally, our boundary descriptions of closed universes indicate that for small but finite $G_N$ there should exist a semiclassical description of a macroscopic closed universe, rather than a one-dimensional Hilbert space, and we discuss weaknesses in arguments favoring the latter.

hep-th↗

arXiv2025

Multicolor Erdős--Rogers Functions

In this paper, we study a multicolor variant of Erdős--Rogers functions. Let $f_{α_s; K_{i_1}, \cdots, K_{i_t}}(n)$ be the largest integer $m$ such that there is always an induced $K_s$-free subgraph of size $m$ in every $n$-vertex graph with a $t$-edge-coloring in which the edges with the $j$-th color induce no copy of $K_{i_j}$. We establish both upper and lower bounds for this multicolor version. Specifically, we show that $f_{α_5; K_3, K_3}(n) = n^{1/2+o(1)}$, $Ω(n^{5/11}) \le f_{α_5; K_3, K_3, K_3}(n) \le n^{1/2+o(1)}$, and $Ω(n^{20/61}) \le f_{α_5; K_3, K_3, K_3, K_3}(n) \le n^{1/3+o(1)}$.

math.CO↗

arXiv2025

H$_{2}$OT: Hierarchical Hourglass Tokenizer for Efficient Video Pose Transformers

Transformers have been successfully applied in the field of video-based 3D human pose estimation. However, the high computational costs of these video pose transformers (VPTs) make them impractical on resource-constrained devices. In this paper, we present a hierarchical plug-and-play pruning-and-recovering framework, called Hierarchical Hourglass Tokenizer (H$_{2}$OT), for efficient transformer-based 3D human pose estimation from videos. H$_{2}$OT begins with progressively pruning pose tokens of redundant frames and ends with recovering full-length sequences, resulting in a few pose tokens in the intermediate transformer blocks and thus improving the model efficiency. It works with two key modules, namely, a Token Pruning Module (TPM) and a Token Recovering Module (TRM). TPM dynamically selects a few representative tokens to eliminate the redundancy of video frames, while TRM restores the detailed spatio-temporal information based on the selected tokens, thereby expanding the network output to the original full-length temporal resolution for fast inference. Our method is general-purpose: it can be easily incorporated into common VPT models on both seq2seq and seq2frame pipelines while effectively accommodating different token pruning and recovery strategies. In addition, our H$_{2}$OT reveals that maintaining the full pose sequence is unnecessary, and a few pose tokens of representative frames can achieve both high efficiency and estimation accuracy. Extensive experiments on multiple benchmark datasets demonstrate both the effectiveness and efficiency of the proposed method. Code and models are available at https://github.com/NationalGAILab/HoT.

cs.CV↗

arXiv2025

The Making of von Neumann Algebras from Bulk Focusing

The single-trace, infinite-N algebra of an arbitrary region may or may not be a von Neumann algebra depending on the GNS sector. In this paper we identify the holographic dual of this mechanism as a consequence of the focusing of null geodesics; more precisely, this GNS sector-dependence corresponds to the well-known difference between null congruences fired from the bulk and those fired from the boundary. As part of establishing this property, we give a rigorous formulation and proof of causal wedge reconstruction for those general boundary subregions whose single trace algebras support von Neumann algebras at large-N. We discuss a possible finite-N extension and interpretation of our results as an explanation for the Hawking area theorem.

hep-th↗

arXiv2025

An edge-spectral Erdős-Stone-Simonovits theorem and its stability

We study the extremal problem that relates the spectral radius $λ(G)$ of an $F$-free graph $G$ with its number of edges. Firstly, we prove that for any graph $F$ with chromatic number $χ(F)=r+1\ge 3$, if $G$ is an $F$-free graph on $m$ edges, then $λ^2(G)\le {(1-\frac{1}{r} + o(1))2m}$. This provides a unified extension of both the Erdős--Stone--Simonovits theorem and its vertex-spectral version due to Nikiforov, and confirms a conjecture proposed by Li, Liu and Feng. We also establish the corresponding edge-spectral stability, showing that if $G$ is an $F$-free graph on $m$ edges with $λ^2(G)=(1- \frac{1}{r} - o(1))2m$, then $G$ differs from a complete bipartite graph by $o(m)$ edges when $r=2$, and $G$ differs from an $r$-partite Turán graph by $o(m)$ edges when $r\ge 3$. This extends the classical Erdős--Simonovits stability theorem. As an application of our method, we improve a result of Zhai, Lin and Shu by showing that if $λ(G)>\sqrt{m}$, then there exist two vertices in $G$ that have at least $\frac{1}{2}\sqrt{m} - O(1)$ common neighbors. This bound is the best possible as witnessed by a random construction.

math.CO↗

arXiv2025

UST-SSM: Unified Spatio-Temporal State Space Models for Point Cloud Video Modeling

Point cloud videos capture dynamic 3D motion while reducing the effects of lighting and viewpoint variations, making them highly effective for recognizing subtle and continuous human actions. Although Selective State Space Models (SSMs) have shown good performance in sequence modeling with linear complexity, the spatio-temporal disorder of point cloud videos hinders their unidirectional modeling when directly unfolding the point cloud video into a 1D sequence through temporally sequential scanning. To address this challenge, we propose the Unified Spatio-Temporal State Space Model (UST-SSM), which extends the latest advancements in SSMs to point cloud videos. Specifically, we introduce Spatial-Temporal Selection Scanning (STSS), which reorganizes unordered points into semantic-aware sequences through prompt-guided clustering, thereby enabling the effective utilization of points that are spatially and temporally distant yet similar within the sequence. For missing 4D geometric and motion details, Spatio-Temporal Structure Aggregation (STSA) aggregates spatio-temporal features and compensates. To improve temporal interaction within the sampled sequence, Temporal Interaction Sampling (TIS) enhances fine-grained temporal dependencies through non-anchor frame utilization and expanded receptive fields. Experimental results on the MSR-Action3D, NTU RGB+D, and Synthia 4D datasets validate the effectiveness of our method. Our code is available at https://github.com/wangzy01/UST-SSM.

cs.CV↗

arXiv2025

More on Nosal's spectral theorem: Books and $4$-cycles

Spectral graph theory studies how the eigenvalues of a graph relate to the structural properties of a graph. In this paper, we solve three open problems in spectral extremal graph theory which generalize the classical Turán-type supersaturation results. (a) We prove that every $m$-edge graph $G$ with the spectral radius $λ(G) > \sqrt{m}$ contains at least $\frac{1}{144} \sqrt{m}$ triangles sharing a common edge. This result confirms a conjecture of Nikiforov, and Li and Peng. Moreover, the bound is optimal up to a constant factor. (b) Next, for $m$-edge graph $G$ with $λ(G) > \sqrt{(1-\frac{1}{r})2m}$, we show that it must contain $Ω_r (\sqrt{m})$ copies of $K_{r+1}$ sharing $r$ common vertices. This confirms a conjecture of Li, Liu and Feng and unifies a series of spectral extremal results on books and cliques. Moreover, we also show that such a graph $G$ contains $Ω_r (m^{\frac{r-1}{2}})$ copies of $K_{r+1}$. This extends a result of Ning and Zhai for counting triangles. (c) We prove that every $m$-edge graph $G$ with $λ(G) > \sqrt{m}$ contains at least $(\frac{1}{8}-o(1)) m^2$ copies of 4-cycles, and we provide two constructions showing that the constant $\frac{1}{8}$ is the best possible. This result settles a problem raised by Ning and Zhai, and it gives the first asymptotics for counting degenerate bipartite graphs. The key to our proof are two structural results we obtain for graphs with large spectral radii on their maximum degree and on existence of large structured subgraphs, which we believe to be of independent interest.

math.CO↗

arXiv2025

Geometric constructions for Ramsey-Turán theory

Combining two classical notions in extremal combinatorics, the study of Ramsey-Turán theory seeks to determine, for integers $m\le n$ and $p \leq q$, the number $\mathsf{RT}_p(n,K_q,m)$, which is the maximum size of an $n$-vertex $K_q$-free graph in which every set of at least $m$ vertices contains a $K_p$. Two major open problems in this area from the 80s ask: (1) whether the asymptotic extremal structure for the general case exhibits certain periodic behaviour, resembling that of the special case when $p=2$; (2) constructing analogues of Bollobás-Erdős graphs with densities other than $1/2$. We refute the first conjecture by witnessing asymptotic extremal structures that are drastically different from the $p=2$ case, and address the second problem by constructing Bollobás-Erdős-type graphs using high dimensional complex spheres with all rational densities. Some matching upper bounds are also provided.

math.CO↗

arXiv2025

Masked Clustering Prediction for Unsupervised Point Cloud Pre-training

Vision transformers (ViTs) have recently been widely applied to 3D point cloud understanding, with masked autoencoding as the predominant pre-training paradigm. However, the challenge of learning dense and informative semantic features from point clouds via standard ViTs remains underexplored. We propose MaskClu, a novel unsupervised pre-training method for ViTs on 3D point clouds that integrates masked point modeling with clustering-based learning. MaskClu is designed to reconstruct both cluster assignments and cluster centers from masked point clouds, thus encouraging the model to capture dense semantic information. Additionally, we introduce a global contrastive learning mechanism that enhances instance-level feature learning by contrasting different masked views of the same point cloud. By jointly optimizing these complementary objectives, i.e., dense semantic reconstruction, and instance-level contrastive learning. MaskClu enables ViTs to learn richer and more semantically meaningful representations from 3D point clouds. We validate the effectiveness of our method via multiple 3D tasks, including part segmentation, semantic segmentation, object detection, and classification, where MaskClu sets new competitive results. The code and models will be released at:https://github.com/Amazingren/maskclu.

cs.CV↗

arXiv2025

SimLabel: Similarity-Weighted Iterative Framework for Multi-annotator Learning with Missing Annotations

Multi-annotator learning (MAL) aims to model annotator-specific labeling patterns. However, existing methods face a critical challenge: they simply skip updating annotator-specific model parameters when encountering missing labels, i.e., a common scenario in real-world crowdsourced datasets where each annotator labels only small subsets of samples. This leads to inefficient data utilization and overfitting risks. To this end, we propose a novel similarity-weighted semi-supervised learning framework (SimLabel) that leverages inter-annotator similarities to generate weighted soft labels for missing annotations, enabling the utilization of unannotated samples rather than skipping them entirely. We further introduce a confidence-based iterative refinement mechanism that combines maximum probability with entropy-based uncertainty to prioritize predicted high-quality pseudo-labels to impute missing labels, jointly enhancing similarity estimation and model performance over time. For evaluation, we contribute a new multimodal multi-annotator dataset, AMER2, with high and more variable missing rates, reflecting real-world annotation sparsity and enabling evaluation across different sparsity levels.

cs.MM↗

arXiv2025

QuMAB: Query-based Multi-Annotator Behavior Modeling with Reliability under Sparse Labels

Multi-annotator learning traditionally aggregates diverse annotations to approximate a single ground truth, treating disagreements as noise. However, this paradigm faces fundamental challenges: subjective tasks often lack absolute ground truth, and sparse annotation coverage makes aggregation statistically unreliable. We introduce a paradigm shift from sample-wise aggregation to annotator-wise behavior modeling. By treating annotator disagreements as valuable information rather than noise, modeling annotator-specific behavior patterns can reconstruct unlabeled data to reduce annotation cost, enhance aggregation reliability, and explain annotator decision behavior. To this end, we propose QuMAB (Query-based Multi-Annotator Behavior Pattern Learning), which uses light-weight queries to model individual annotators while capturing inter-annotator correlations as implicit regularization, preventing overfitting to sparse individual data while maintaining individualization and improving generalization, with a visualization of annotator focus regions offering an explainable analysis of behavior understanding. We contribute two large-scale datasets with dense per-annotator labels: STREET (4,300 labels/annotator) and AMER (average 3,118 labels/annotator), the first multimodal multi-annotator dataset. Extensive experiments demonstrate the superiority of our QuMAB in modeling individual annotators' behavior patterns, their utility for consensus prediction, and applicability under sparse annotations.

cs.MM↗

arXiv2025

Volume as an index of a subalgebra

We propose a new way to understand the volume of certain subregions in the bulk of AdS spacetime by relating it to an algebraic quantity known as the index of inclusion. This index heuristically measures the relative size of a subalgebra $\mathcal{N}$ embedded within a larger algebra $\mathcal{M}$. According to subregion-subalgebra duality, bulk subregions are described by von Neumann algebras on the boundary. When a causally complete bulk subregion corresponds to the relative commutant $\mathcal{N}' \cap \mathcal{M}$ -- the set of operators in $\mathcal{M}$ that commute with $\mathcal{N}$ -- of boundary subalgebras, we propose that the exponential of the volume of the maximal volume slice of the subregion equals the index of inclusion. This ``volume-index'' relation provides a new boundary explanation for the growth of interior volume in black holes, reframing it as a change in the relative size of operator algebras. It offers a complementary perspective on complexity growth from the Heisenberg picture, and has a variety of other applications, including quantifying the relative size of algebras dual to the entanglement wedge and the causal wedge of a boundary region, as well as quantifying the violation of additivity of operator algebras in the large $N$ limit. Finally, it may offer insights into the volume growth of de Sitter space through the changes in North and South pole observer algebras in time.

hep-th↗

arXiv2025

Recognizing Actions from Robotic View for Natural Human-Robot Interaction

Natural Human-Robot Interaction (N-HRI) requires robots to recognize human actions at varying distances and states, regardless of whether the robot itself is in motion or stationary. This setup is more flexible and practical than conventional human action recognition tasks. However, existing benchmarks designed for traditional action recognition fail to address the unique complexities in N-HRI due to limited data, modalities, task categories, and diversity of subjects and environments. To address these challenges, we introduce ACTIVE (Action from Robotic View), a large-scale dataset tailored specifically for perception-centric robotic views prevalent in mobile service robots. ACTIVE comprises 30 composite action categories, 80 participants, and 46,868 annotated video instances, covering both RGB and point cloud modalities. Participants performed various human actions in diverse environments at distances ranging from 3m to 50m, while the camera platform was also mobile, simulating real-world scenarios of robot perception with varying camera heights due to uneven ground. This comprehensive and challenging benchmark aims to advance action and attribute recognition research in N-HRI. Furthermore, we propose ACTIVE-PC, a method that accurately perceives human actions at long distances using Multilevel Neighborhood Sampling, Layered Recognizers, Elastic Ellipse Query, and precise decoupling of kinematic interference from human actions. Experimental results demonstrate the effectiveness of ACTIVE-PC. Our code is available at: https://github.com/wangzy01/ACTIVE-Action-from-Robotic-View.

cs.CV↗