arXiv subjects

Zhen Zhang

Publications and source records attributed to Zhen Zhang.

At least 199 records · Page 11Linked to original sources

arXiv2025

WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning

Transcending human cognitive limitations represents a critical frontier in LLM training. Proprietary agentic systems like DeepResearch have demonstrated superhuman capabilities on extremely complex information-seeking benchmarks such as BrowseComp, a feat previously unattainable. We posit that their success hinges on a sophisticated reasoning pattern absent in open-source models: the ability to systematically reduce extreme uncertainty when navigating vast information landscapes. Based on this insight, we introduce WebSailor, a complete post-training methodology designed to instill this crucial capability. Our approach involves generating novel, high-uncertainty tasks through structured sampling and information obfuscation, RFT cold start, and an efficient agentic RL training algorithm, Duplicating Sampling Policy Optimization (DUPO). With this integrated pipeline, WebSailor significantly outperforms all open-source agents in complex information-seeking tasks, matching proprietary agents' performance and closing the capability gap.

cs.LG↗

arXiv2025

Scaling Agents via Continual Pre-training

Large language models (LLMs) have evolved into agentic systems capable of autonomous tool use and multi-step reasoning for complex problem-solving. However, post-training approaches building upon general-purpose foundation models consistently underperform in agentic tasks, particularly in open-source implementations. We identify the root cause: the absence of robust agentic foundation models forces models during post-training to simultaneously learn diverse agentic behaviors while aligning them to expert demonstrations, thereby creating fundamental optimization tensions. To this end, we are the first to propose incorporating Agentic Continual Pre-training (Agentic CPT) into the deep research agents training pipeline to build powerful agentic foundational models. Based on this approach, we develop a deep research agent model named AgentFounder. We evaluate our AgentFounder-30B on 10 benchmarks and achieve state-of-the-art performance while retains strong tool-use ability, notably 39.9% on BrowseComp-en, 43.3% on BrowseComp-zh, and 31.5% Pass@1 on HLE.

cs.CL↗

arXiv2025

Towards General Agentic Intelligence via Environment Scaling

Advanced agentic intelligence is a prerequisite for deploying Large Language Models in practical, real-world applications. Diverse real-world APIs demand precise, robust function-calling intelligence, which needs agents to develop these capabilities through interaction in varied environments. The breadth of function-calling competence is closely tied to the diversity of environments in which agents are trained. In this work, we scale up environments as a step towards advancing general agentic intelligence. This gives rise to two central challenges: (i) how to scale environments in a principled manner, and (ii) how to effectively train agentic capabilities from experiences derived through interactions with these environments. To address these, we design a scalable framework that automatically constructs heterogeneous environments that are fully simulated, systematically broadening the space of function-calling scenarios. We further adapt a two-phase agent fine-tuning strategy: first endowing agents with fundamental agentic capabilities, then specializing them for domain-specific contexts. Extensive experiments on agentic benchmarks, tau-bench, tau2-Bench, and ACEBench, demonstrate that our trained model, AgentScaler, significantly enhances the function-calling capability of models.

cs.CL↗

arXiv2025

Verifying Computational Graphs in Production-Grade Distributed Machine Learning Frameworks

Modern machine learning frameworks support very large models by incorporating parallelism and optimization techniques. Yet, these very techniques add new layers of complexity, introducing silent errors that severely degrade model performance. Existing solutions are either ad hoc or too costly for production. We present Scalify, a lightweight framework that exposes silent errors by verifying semantic equivalence of computational graphs using equality saturation and Datalog-style reasoning. To scale, Scalify partitions graphs with parallel rewriting and layer memoization, reuses rewrite templates, and augments equality saturation with relational reasoning and symbolic bijection inference. It further localizes discrepancies to precise code sites, turning verification results into actionable debugging guidance. Scalify verifies models as large as Llama-3.1-405B within minutes on a commodity machine and exposed five unknown bugs in Amazon production machine learning frameworks.

cs.LG↗

arXiv2025

Theory meets experiment: insights into structure and magnetic properties of Fe$_{1-x}$Ni$_{x}$B alloy

We studied the structural and magnetic properties of the solid solution Fe$_{1-x}$Ni$_{x}$B through theoretical and experimental approaches. Powder X-ray diffraction, X-ray Pair Distribution Function analysis, and energy dispersive X-ray spectroscopy reveal that the Fe$_{1-x}$Ni$_{x}$B solid solution crystallizes in the $β$-FeB structure type up to x = 0.6-0.7 and exhibits anisotropic unit cell volume contraction with increasing Ni concentration. Magnetic measurements showed a transition from ferromagnetism to paramagnetism around x = 0.7. For x = 0.5, the low (< 0.3 $μ_{B}$) magnetic moments suggest itinerant magnetism despite the relatively high Curie temperature (up to 225 K). Theoretical calculations indicated different types of magnetic orderings depending on the Fe/Ni atomic order, with the antiferromagnetic state being stable for ordered FeNiB$_{2}$, whereas the ground state is ferromagnetic for the disordered alloy. Calculations also predicted the coexistence of low- and high-spin states in Fe atoms around the composition with x=0.5, in line with the experimental evidence from $^{57}$Fe Mössbauer spectroscopy. The two magnetically distinct Fe sites for x = 0.3, 0.4, and 0.5 observed by $^{57}$Fe Mössbauer spectroscopy can also be interpreted as two magnetically different regions or clusters that could affect the critical behavior near a quantum magnetic transition based on a potential ferromagnetic quantum critical point identified computationally and experimentally near x=0.64. This work highlights the complex interplay between structure and magnetism in Fe$_{1-x}$Ni$_{x}$B alloys, suggesting areas for future research on quantum critical behavior.

cond-mat.mtrl-sci↗

arXiv2025

Digital Twin Channel-Aided CSI Prediction: An Environment-Based Subspace Extraction Approach for Achieving Low Overhead and High Robustness

To meet the robust and high-speed communication requirements of the sixth-generation (6G) mobile communication system in complex scenarios, sensing- and artificial intelligence (AI)-based digital twin channel (DTC) techniques become a promising approach to reduce system overhead. In this paper, we propose an environment-specific channel subspace basis (ECB)-aided partial-to-whole channel state information (CSI) prediction method (ECB-P2WCP) for realizing DTC-enabled low-overhead channel prediction. Specifically, we introduce a wireless environment knowledge (WEK) construction method that extracts ECB from the digital twin environment via subspace estimation. This ECB characterizes the static statistical properties of the electromagnetic environment and serves as environment information prior to the prediction task. Then, we fuse ECB with real-time estimated local CSI to predict the entire spatial-frequency domain channel for both the present and future time instances. Hence, an ECB-based partial-to-whole CSI prediction network (ECB-P2WNet) is designed to achieve a robust channel prediction scheme in various complex scenarios. Simulation results indicate that incorporating ECB provides significant benefits under low signal-to-noise ratio and pilot ratio conditions, achieving a reduction of up to 50\% in pilot overhead. Additionally, the proposed method maintains robustness against multi-user interference, tolerating 3-meter localization errors with only a 0.5 dB normalized mean square error increase, and predicts CSI for the next channel coherent time within 1.3 milliseconds.

eess.SP↗

arXiv2025

Prospects for Time-Domain and Multi-Messenger Science with eXTP

In this new era of time-domain and multi-messenger astronomy, various new transients and new phenomena are constantly being discovered thanks to the rapid advances in observations, which provide the excellent opportunity to study the physics in the extreme environments. The enhanced X-ray Timing and Polarimetry mission (eXTP), planned to be launched in 2030, has several key advantages, including advanced polarimetry, high sensitivity & large effective area, and wide energy range coverage, which make it a groundbreaking project in high-energy astrophysics. In this article, we briefly introduce the potential time-domain and multi-messenger targets for eXTP, including gravitational-wave (GW) counterparts, gamma-ray bursts (GRBs), magnetars and fast radio bursts (FRBs), tidal disruption events (TDEs), supernovae, high energy neutrinos and TeV active galactic nucleus (AGNs), and so on. We discuss the advantages of future eXTP observations for detecting these sources, their detection capabilities, the abilities to distinguish theoretical models, and their applications in gravity and cosmology.

astro-ph.HE↗

arXiv2025

Language Model Perplexity Predicts Scientific Surprise and Transformative Impact

Scientific breakthroughs typically emerge through the surprising violation of established research ideas, yet quantifying surprise has remained elusive because it requires a coherent model of all contemporary scientific worldviews. Deep neural networks like large language models (LLMs) are arbitrary function approximators tuned to consistently expect the expressions and ideas on which they were trained and those semantically nearby. This suggests that as LLMs improve at generating plausible text, so the perplexity or improbability a text sequence would be generated by them should come to better predict scientific surprise and disruptive importance. Analyzing over 2 million papers across multiple disciplines published immediately following the training of 5 prominent open LLMs, here we show that higher perplexity scores systematically predict papers that receive more variable review ratings, longer editorial delays, and greater reviewer uncertainty. The most perplexing papers exhibit bimodal outcomes: disproportionately represented among the most celebrated scientific achievements and also the most discounted. High-perplexity papers tend to be published in journals with more variable impact factors and receive fewer short-term citations but in prestigious venues that bet on long-term impact. They also generate more interdisciplinary engagement portending long-term influence, and are more likely to have been supported by speculative funders like DARPA versus the NIH. Interestingly, we find the opposite pattern for humanities research, where the least surprising work is the most celebrated and cited. Our findings reveal that computational measures of corpus-wide linguistic surprise can forecast the reception and ultimate influence of scientific ideas, offering a scalable approach to recognize and generate potentially transformative research that challenge conventional scientific thinking.

cs.SI↗

arXiv2025

WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agent

Web agents such as Deep Research have demonstrated superhuman cognitive abilities, capable of solving highly challenging information-seeking problems. However, most research remains primarily text-centric, overlooking visual information in the real world. This makes multimodal Deep Research highly challenging, as such agents require much stronger reasoning abilities in perception, logic, knowledge, and the use of more sophisticated tools compared to text-based agents. To address this limitation, we introduce WebWatcher, a multi-modal Agent for Deep Research equipped with enhanced visual-language reasoning capabilities. It leverages high-quality synthetic multimodal trajectories for efficient cold start training, utilizes various tools for deep reasoning, and further enhances generalization through reinforcement learning. To better evaluate the capabilities of multimodal agents, we propose BrowseComp-VL, a benchmark with BrowseComp-style that requires complex information retrieval involving both visual and textual information. Experimental results show that WebWatcher significantly outperforms proprietary baseline, RAG workflow and open-source agents in four challenging VQA benchmarks, which paves the way for solving complex multimodal information-seeking tasks.

cs.IR↗

arXiv2025

Rigidity of Circle Packings with Flexible Radii

Circle packings are arrangement of circles satisfying specified tangency requirements. Many problems about packing of circles and spheres occur in nature particularly in material design and protein structure. Surprisingly, little is known about the stability and rigidity of circle packings. In this paper, we study the rigidity of circle packings representing a given planar graph. The radii of circles are flexible with equality and inequality constraints. We provide a dual condition for the packing to be rigid in the first order. This gives us a sufficient condition to show a packing is rigid. Then we will explore the difficulties on rigidity problems beyond the first order.

math.MG↗

arXiv2025

Detecting Knowledge Boundary of Vision Large Language Models by Sampling-Based Inference

Despite the advancements made in Vision Large Language Models (VLLMs), like text Large Language Models (LLMs), they have limitations in addressing questions that require real-time information or are knowledge-intensive. Indiscriminately adopting Retrieval Augmented Generation (RAG) techniques is an effective yet expensive way to enable models to answer queries beyond their knowledge scopes. To mitigate the dependence on retrieval and simultaneously maintain, or even improve, the performance benefits provided by retrieval, we propose a method to detect the knowledge boundary of VLLMs, allowing for more efficient use of techniques like RAG. Specifically, we propose a method with two variants that fine-tune a VLLM on an automatically constructed dataset for boundary identification. Experimental results on various types of Visual Question Answering datasets show that our method successfully depicts a VLLM's knowledge boundary, based on which we are able to reduce indiscriminate retrieval while maintaining or improving the performance. In addition, we show that the knowledge boundary identified by our method for one VLLM can be used as a surrogate boundary for other VLLMs. Code will be released at https://github.com/Chord-Chen-30/VLLM-KnowledgeBoundary

cs.CL↗

arXiv2025

Discovery and Synthesis of a Family of Boride Altermagnets

Borides are a rich material family. To push the boundaries of borides' properties and applications into broader fields, we have conducted systematic theoretical and experimental searches for synthesizable phases in ternary borides TM$_2$B$_2$ (T = 3d, M = 4d/5d transition metals). We find that TM$_2$B$_2$ in the FeMo$_2$B$_2$-type and CoW$_2$B$_2$-type structures form a large family of stable/metastable materials of 120 members. Among them, we identify 40 materials with stable magnetic solutions. Further, we discover 11 altermagnets in the FeMo$_2$B$_2$-type structure. So far, boride altermagnets are rare. In these altermagnets, T = Fe or Mn atoms are arranged in parallel T-chains with strong ferromagnetic intrachain couplings and antiferromagnetic interchain couplings. They simultaneously exhibit electronic band spin splitting, typical of ferromagnetism, and zero net magnetization, typical of antiferromagnetism. They also exhibit magnonic band chiral splitting. Both effects originate from the unique altermagnetic symmetries crucially constrained by the nonmagnetic atoms in the structure. Transport properties of relevance to spintronic applications, including the strain-induced spin-splitter effect and anomalous Hall effect, are predicted. An iodine-assisted synthesis method for TM$_2$B$_2$ is developed, using which 7 of the predicted low-energy phases are experimentally synthesized and characterized, including 4 altermagnets. This work expands the realm of borides by offering new opportunities for studying altermagnetism and altermagnons in borides. It also provides valuable insights into the discovery and design of altermagnets. By demonstrating that altermagnets can exist as families sharing a common motif, this work paves a feasible route for discovering altermagnets by elemental substitutions and high-throughput computations.

cond-mat.mtrl-sci↗

arXiv2025

Onsager Principle-Based Domain Embedding for Thermodynamically Consistent Cahn-Hilliard Model in Arbitrary Domain

The original Cahn-Hilliard model in an arbitrary domain with two prescribed boundary conditions is extended to a Cahn-Hilliard-type model in a larger, regular domain with homogeneous Neumann boundary conditions. The extension is based on the Onsager principle-based domain embedding (OPBDE) method, which has been developed as a systematic domain embedding framework to ensure thermodynamic consistency. By introducing a modified conservation law, the flux at the boundary of the original domain is incorporated into the conservation law as a source term. Our variational approach demonstrates that, even without a prior knowledge on the specific form of the rate of free energy pumped into the system, the Onsager principle remains an effective instrument in deriving the constitutive equation of the extended system. This approach clarifies the intrinsic structure of the extended model in the perspectives of free energy and its dissipation. Asymptotic analysis is carried out for the extended OPBDE Cahn-Hilliard model, demonstrating that the original Cahn-Hilliard model, including its boundary conditions, can be fully recovered. To validate our approach, a structure-preserving numerical scheme is developed to discretize the extended model. Numerical results show that the OPBDE Cahn-Hilliard model is accurate, effective, and robust, highlighting the capability of the OPBDE method in handling gradient flow problems in arbitrary domain geometries.

math.NA↗

arXiv2025

Causal Disentanglement and Cross-Modal Alignment for Enhanced Few-Shot Learning

Few-shot learning (FSL) often requires effective adaptation of models using limited labeled data. However, most existing FSL methods rely on entangled representations, requiring the model to implicitly recover the unmixing process to obtain disentangled representations using only limited supervision, which hinders effective adaptation. Recent theoretical studies show that multimodal contrastive learning methods, such as CLIP, can disentangle latent representations up to linear transformations. In light of this, we propose the Causal CLIP Adapter (CCA), a novel framework that explicitly disentangles visual features extracted from CLIP using unsupervised Independent Component Analysis (ICA). This removes the need to learn the unmixing process from the labeled data, thereby reducing the number of trainable parameters and mitigating overfitting. Taking a step further, while ICA can obtain visual disentangled representations, it may also disrupt CLIP's intra- and inter-modal alignment. To counteract this, CCA further leverages CLIP's inherent cross-modal alignment by enhancing it in two ways: unidirectionally, through fine-tuning a CLIP-based text classifier, and bidirectionally, via a cross-attention mechanism that enriches visual and textual representations through mutual interaction. Both unimodal and cross-modal classification outputs can be effectively combined linearly to improve classification accuracy. Extensive experiments on 11 benchmark datasets demonstrate that our method consistently outperforms state-of-the-art approaches in terms of few-shot performance and robustness to distributional shifts, while maintaining computational efficiency. Code will be available at https://github.com/tianjiao-j/CCA.

cs.CV↗

arXiv2025

Efficient Federated Learning with Encrypted Data Sharing for Data-Heterogeneous Edge Devices

As privacy protection gains increasing importance, more models are being trained on edge devices and subsequently merged into the central server through Federated Learning (FL). However, current research overlooks the impact of network topology, physical distance, and data heterogeneity on edge devices, leading to issues such as increased latency and degraded model performance. To address these issues, we propose a new federated learning scheme on edge devices that called Federated Learning with Encrypted Data Sharing(FedEDS). FedEDS uses the client model and the model's stochastic layer to train the data encryptor. The data encryptor generates encrypted data and shares it with other clients. The client uses the corresponding client's stochastic layer and encrypted data to train and adjust the local model. FedEDS uses the client's local private data and encrypted shared data from other clients to train the model. This approach accelerates the convergence speed of federated learning training and mitigates the negative impact of data heterogeneity, making it suitable for application services deployed on edge devices requiring rapid convergence. Experiments results show the efficacy of FedEDS in promoting model performance.

cs.LG↗

arXiv2025

LITE: A Learning-Integrated Topological Explorer for Multi-Floor Indoor Environments

This work focuses on multi-floor indoor exploration, which remains an open area of research. Compared to traditional methods, recent learning-based explorers have demonstrated significant potential due to their robust environmental learning and modeling capabilities, but most are restricted to 2D environments. In this paper, we proposed a learning-integrated topological explorer, LITE, for multi-floor indoor environments. LITE decomposes the environment into a floor-stair topology, enabling seamless integration of learning or non-learning-based 2D exploration methods for 3D exploration. As we incrementally build floor-stair topology in exploration using YOLO11-based instance segmentation model, the agent can transition between floors through a finite state machine. Additionally, we implement an attention-based 2D exploration policy that utilizes an attention mechanism to capture spatial dependencies between different regions, thereby determining the next global goal for more efficient exploration. Extensive comparison and ablation studies conducted on the HM3D and MP3D datasets demonstrate that our proposed 2D exploration policy significantly outperforms all baseline explorers in terms of exploration efficiency. Furthermore, experiments in several 3D multi-floor environments indicate that our framework is compatible with various 2D exploration methods, facilitating effective multi-floor indoor exploration. Finally, we validate our method in the real world with a quadruped robot, highlighting its strong generalization capabilities.

cs.RO↗

arXiv2025

Constraining neutron-proton effective mass splitting through nuclear giant dipole resonance within transport approach

Based on the Boltzmann-Uehling-Uhlenbeck equation, we investigate the effects of the isovector nucleon effective mass $m^*_{v,0}$ and the in-medium nucleon-nucleon cross section $σ^*$ on the isovector giant dipole resonance~(IVGDR) in $^{208}{\rm Pb}$, employing a set of representative Skyrme energy density functionals. We find that the energy-weighted sum rule $m_1$ of the IVGDR is highly sensitive to $m^{*}_{v,0}$ and only mildly dependent on $σ^*$, while the width $Γ$ of the IVGDR is primarily governed by $σ^*$ with a moderate sensitivity to $m^*_{v,0}$. From a Bayesian analysis of both $m_1$ and $Γ$, we infer the isovector effective mass $m^{*}_{v,0}/m$ = $0.731^{+0.027}_{-0.023}$, where $m$ is the bare nucleon mass. Furthermore, by incorporating the isoscalar effective mass $m^*_{s,0}/m = 0.820 \pm 0.030$, extracted from the isoscalar giant quadrupole resonance in $^{208}{\rm Pb}$, the linear neutron-proton effective mass splitting coefficient at saturation density $ρ_0$ is determined to be $Δm^*_1 (ρ_0)/m = 0.200 ^{+0.101}_{-0.094}$.

nucl-th↗

arXiv2025

Behavior-Specific Filtering for Enhanced Pig Behavior Classification in Precision Livestock Farming

This study proposes a behavior-specific filtering method to improve behavior classification accuracy in Precision Livestock Farming. While traditional filtering methods, such as wavelet denoising, achieved an accuracy of 91.58%, they apply uniform processing to all behaviors. In contrast, the proposed behavior-specific filtering method combines Wavelet Denoising with a Low Pass Filter, tailored to active and inactive pig behaviors, and achieved a peak accuracy of 94.73%. These results highlight the effectiveness of behavior-specific filtering in enhancing animal behavior monitoring, supporting better health management and farm efficiency.

cs.LG↗