arXiv subjects

Zhao Xue

Publications and source records attributed to Zhao Xue.

12 recordsLinked to original sources

arXiv2026

GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal Agents

We present GLM-5V-Turbo, a step toward native foundation models for multimodal agents. As foundation models are increasingly deployed in real environments, agentic capability depends not only on language reasoning, but also on the ability to perceive, interpret, and act over heterogeneous contexts such as images, videos, webpages, documents, GUIs. GLM-5V-Turbo is built around this objective: multimodal perception is integrated as a core component of reasoning, planning, tool use, and execution, rather than as an auxiliary interface to a language model. This report summarizes the main improvements behind GLM-5V-Turbo across model design, multimodal training, reinforcement learning, toolchain expansion, and integration with agent frameworks. These developments lead to strong performance in multimodal coding, visual tool use, and framework-based agentic tasks, while preserving competitive text-only coding capability. More importantly, our development process offers practical insights for building multimodal agents, highlighting the central role of multimodal perception, hierarchical optimization, and reliable end-to-end verification.

cs.CV↗

arXiv2025

GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

We present GLM-4.1V-Thinking, GLM-4.5V, and GLM-4.6V, a family of vision-language models (VLMs) designed to advance general-purpose multimodal understanding and reasoning. In this report, we share our key findings in the development of the reasoning-centric training framework. We first develop a capable vision foundation model with significant potential through large-scale pre-training, which arguably sets the upper bound for the final performance. We then propose Reinforcement Learning with Curriculum Sampling (RLCS) to unlock the full potential of the model, leading to comprehensive capability enhancement across a diverse range of tasks, including STEM problem solving, video understanding, content recognition, coding, grounding, GUI-based agents, and long document interpretation. In a comprehensive evaluation across 42 public benchmarks, GLM-4.5V achieves state-of-the-art performance on nearly all tasks among open-source models of similar size, and demonstrates competitive or even superior results compared to closed-source models such as Gemini-2.5-Flash on challenging tasks including Coding and GUI Agents. Meanwhile, the smaller GLM-4.1V-9B-Thinking remains highly competitive-achieving superior results to the much larger Qwen2.5-VL-72B on 29 benchmarks. We open-source both GLM-4.1V-9B-Thinking and GLM-4.5V. We further introduce the GLM-4.6V series, open-source multimodal models with native tool use and a 128K context window. A brief overview is available at https://z.ai/blog/glm-4.6v. Code, models and more information are released at https://github.com/zai-org/GLM-V.

cs.CV↗

arXiv2024

CogVLM2: Visual Language Models for Image and Video Understanding

Beginning with VisualGLM and CogVLM, we are continuously exploring VLMs in pursuit of enhanced vision-language fusion, efficient higher-resolution architecture, and broader modalities and applications. Here we propose the CogVLM2 family, a new generation of visual language models for image and video understanding including CogVLM2, CogVLM2-Video and GLM-4V. As an image understanding model, CogVLM2 inherits the visual expert architecture with improved training recipes in both pre-training and post-training stages, supporting input resolution up to $1344 \times 1344$ pixels. As a video understanding model, CogVLM2-Video integrates multi-frame input with timestamps and proposes automated temporal grounding data construction. Notably, CogVLM2 family has achieved state-of-the-art results on benchmarks like MMBench, MM-Vet, TextVQA, MVBench and VCGBench. All models are open-sourced in https://github.com/THUDM/CogVLM2 and https://github.com/THUDM/GLM-4, contributing to the advancement of the field.

cs.CV↗

arXiv2022

WuDaoMM: A large-scale Multi-Modal Dataset for Pre-training models

Compared with the domain-specific model, the vision-language pre-training models (VLPMs) have shown superior performance on downstream tasks with fast fine-tuning process. For example, ERNIE-ViL, Oscar and UNIMO trained VLPMs with a uniform transformers stack architecture and large amounts of image-text paired data, achieving remarkable results on downstream tasks such as image-text reference(IR and TR), vision question answering (VQA) and image captioning (IC) etc. During the training phase, VLPMs are always fed with a combination of multiple public datasets to meet the demand of large-scare training data. However, due to the unevenness of data distribution including size, task type and quality, using the mixture of multiple datasets for model training can be problematic. In this work, we introduce a large-scale multi-modal corpora named WuDaoMM, totally containing more than 650M image-text pairs. Specifically, about 600 million pairs of data are collected from multiple webpages in which image and caption present weak correlation, and the other 50 million strong-related image-text pairs are collected from some high-quality graphic websites. We also release a base version of WuDaoMM with 5 million strong-correlated image-text pairs, which is sufficient to support the common cross-modal model pre-training. Besides, we trained both an understanding and a generation vision-language (VL) model to test the dataset effectiveness. The results show that WuDaoMM can be applied as an efficient dataset for VLPMs, especially for the model in text-to-image generation task. The data is released at https://data.wudaoai.cn

cs.CV↗

arXiv2014

A possible method for non-Hermitian and non-$PT$-symmetric Hamiltonian systems

A possible method to investigate non-Hermitian Hamiltonians is suggested through finding a Hermitian operator $η_+$ and defining the annihilation and creation operators to be $η_+$-pseudo-Hermitian adjoint to each other. The operator $η_+$ represents the $η_+$-pseudo-Hermiticity of Hamiltonians. As an example, a non-Hermitian and non-$PT$-symmetric Hamiltonian with imaginary linear coordinate and linear momentum terms is constructed and analyzed in detail. The operator $η_+$ is found, based on which, a real spectrum and a positive-definite inner product, together with the probability explanation of wave functions, the orthogonality of eigenstates, and the unitarity of time evolution, are obtained for the non-Hermitian and non-$PT$-symmetric Hamiltonian. Moreover, this Hamiltonian turns out to be coupled when it is extended to the canonical noncommutative space with noncommutative spatial coordinate operators and noncommutative momentum operators as well. Our method is applicable to the coupled Hamiltonian. Then the first and second order noncommutative corrections of energy levels are calculated, and in particular the reality of energy spectra, the positive-definiteness of inner products, and the related properties (the probability explanation of wave functions, the orthogonality of eigenstates, and the unitarity of time evolution) are found not to be altered by the noncommutativity.

quant-ph↗

arXiv2014

Holographic Superconductors in f(R) Gravity

We study the holographic superconductors in f(R) gravity, and show how the critical temperature and the condensate of the dual operator depend on the modifications to Einstein gravity. We first review the black holes with constant curvature, which are clear in f(R) gravity. The dependencies are represented with the modified Newtonian constant and the effective cosmological constant trivially. Then we consider a planar black hole solution for a specific f(R) without imposing the constant curvature, which is asymptotic to AdS spacetime. The corrections to the thermal dynamics of the black holes, the critical temperature and the condensates are all obtained in a perturbative approach. Some comments are given on the affections of the f(R) gravity to the holographic superconductors.

gr-qc↗

arXiv2013

Critical magnetic field in holographic superconductor in Gauss-Bonnet gravity with Born-Infeld electrodynamics

In the paper using matching method in the probe limit, we investigate some properties of holographic superconductor in Gauss-Bonnet gravity with Born-Infeld electrodynamics . We discuss the effects of the Gauss-Bonnet coupling $\a$ and Born-Infeld parameter $b$ on the critical temperature and condensate. We find that both of $\a$ and $b$ make the critical temperature decrease, which implies the condensate harder to form. Moreover we study the magnetic effect on holographic superconductor and obtain that the ratio between the critical magnetic field and the square of the critical temperature increases from zero as the temperature is lowered below the critical value $T_c$, which agrees well with the former results. We also find the critical magnetic field is indeed affected by Gauss-Bonnet coupling, but not by Born-Infeld parameter.

hep-th↗

arXiv2013

Note on Thermodynamics Method of Black Hole/CFT Correspondence

In the paper we further refine the thermodynamics method of black hole/CFT correspondence. We show that one can derive the central charges of different holographic pictures directly from the entropy product $S_+S_-$ if it is mass-independent, for a black hole in the Einstein gravity or the gravity without diffeomorphism anomaly. For a general black hole in the Einstein gravity that admits holographic descriptions, we show that the thermodynamics method and asymptotic symmetry group (ASG) analysis can always give consistent results in the extreme limit. Furthermore, we discuss the relation between black hole thermodynamics and the hidden conformal symmetry. We show that the condition $T_+A_+=T_-A_-$, with $A_\pm$ being the outer and inner horizon areas, is the necessary, but not sufficient, condition for a black hole to have the hidden conformal symmetry. In particular, for the Einstein(-Maxwell) gravity $T_+A_+=T_-A_-$ is just the condition $T_+S_+=T_-S_-$, with $S_\pm$ being the outer and inner horizon entropies, which is the condition for the entropy product $S_+S_-$ being mass-dependent. When there exists the hidden conformal symmetry in the low-frequency scattering off the generic non-extremal black hole, it always leads to the same temperatures of dual CFT as the ones got from the thermodynamics method.

hep-th↗

arXiv2012

Quantum tunneling and spectroscopy of noncommutative inspired Kerr black hole

We discuss the thermodynamics of the noncommutative inspired Kerr black hole by means of a reformulated Hamilton-Jacobi method and a dimensional reduction technique. In order to investigate the effect of the angular momentum of the tunneling particle, we calculate the wave function to the first order of the WKB ansatz. Then, using a density matrix technique we derive the radiation spectrum from which the radiation temperature can be read out. Our results show that the radiation of this noncommutative inspired black hole corresponds to a modified temperature which involves the effect of noncommutativity. However, the angular momentum of the tunneling particle has no influence on the radiation temperature. Moreover, we analyze the entropy spectrum and verify that its quantization is modified neither by the noncommutativity of spacetime nor by the quantum correction of wave functions.

hep-th↗

arXiv2011

Tunneling of massive particles from noncommutative inspired Schwarzschild black hole

We apply the generalization of the Parikh-Wilczek method to the tunneling of massive particles from noncommutative inspired Schwarzschild black holes. By deriving the equation of radial motion of the tunneling particle directly, we calculate the emission rate which is shown to be dependent on the noncommutative parameter besides the energy and mass of the tunneling particle. After equating the emission rate to the Boltzmann factor, we obtain the modified Hawking temperature which relates to the noncommutativity and recovers the standard Hawking temperature in the commutative limit. We also discuss the entropy of the noncommutative inspired Schwarzschild black hole and its difference after and before a massive particle's emission.

hep-th↗

arXiv2011

Massive charged particle's tunneling from spherical charged black hole

We generalize the Parikh-Wilczek scheme to the tunneling of a massive charged particle from a general spherical charged black hole. We obtain that the tunneling probability depends on the energy, the mass and the charge of the particle. In particular, the modified Hawking temperature is related to the charge. Only at the leading order approximation can the standard Hawking temperature be reproduced. We take the Reissner-Nordström black hole as an example to clarify our points of view, and find that the accumulation of Hawking radiation makes it approach an extreme black hole.

hep-th↗

arXiv2011

U(2,2) gravity on noncommutative space with symplectic structure

The classical Einstein's gravity can be reformulated from the constrained U(2,2) gauge theory on the ordinary (commutative) four-dimensional spacetime. Here we consider a noncommutative manifold with a symplectic structure and construct a U(2,2) gauge theory on such a manifold by using the covariant coordinate method. Then we use the Seiberg-Witten map to express noncommutative quantities in terms of their commutative counterparts up to the first-order in noncommutative parameters. After imposing constraints we obtain a noncommutative gravity theory described by the Lagrangian with up to nonvanishing first order corrections in noncommutative parameters. This result coincides with our previous one obtained for the noncommutative SL(2,C) gravity.

hep-th↗