arXiv subjects

Long Zhang

Publications and source records attributed to Long Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2025

Each Prompt Matters: Scaling Reinforcement Learning Without Wasting Rollouts on Hundred-Billion-Scale MoE

We present CompassMax-V3-Thinking, a hundred-billion-scale MoE reasoning model trained with a new RL framework built on one principle: each prompt must matter. Scaling RL to this size exposes critical inefficiencies-zero-variance prompts that waste rollouts, unstable importance sampling over long horizons, advantage inversion from standard reward models, and systemic bottlenecks in rollout processing. To overcome these challenges, we introduce several unified innovations: (1) Multi-Stage Zero-Variance Elimination, which filters out non-informative prompts and stabilizes group-based policy optimization (e.g. GRPO) by removing wasted rollouts; (2) ESPO, an entropy-adaptive optimization method that balances token-level and sequence-level importance sampling to maintain stable learning dynamics; (3) a Router Replay strategy that aligns training-time MoE router decisions with inference-time behavior to mitigate train-infer discrepancies, coupled with a reward model adjustment to prevent advantage inversion; (4) a high-throughput RL system with FP8-precision rollouts, overlapped reward computation, and length-aware scheduling to eliminate performance bottlenecks. Together, these contributions form a cohesive pipeline that makes RL on hundred-billion-scale MoE models stable and efficient. The resulting model delivers strong performance across both internal and public evaluations.

cs.AI↗

arXiv2025

Dimension- and Facet-Dependent Altermagnetic Biferroics and Ferromagnetic Biferroics and Triferroics in CrSb

Altermagnets have recently garnered significant interest due to their vanishing net magnetic moment and non-relativistic momentum-dependent spin splitting. However, altermagnetic (AM) multiferroics especially triferroics remain scarce. We investigate the experimentally synthesized non-van der Waals CrSb as a model system to explore the effects of dimensionality and facet orientation on its ferroic properties. NiAs, MnP, wurtzite (WZ), zincblende (ZB), and rocksalt (RS) phases are considered. Using first-principles calculations, we predict the altermagnetism of CrSb in MnP phase which has comparable stability with experimental NiAs phase. Both NiAs- and MnP-phase (110) facets exhibit AM-ferroelastic (FC) biferroics, while the WZ-phase bulk and (001) facets host ferromagnetic (FM) or AM-ferroelectric (FE) biferroics. Notably, the WZ-phase (110) facets are identified as FM/AM-FE-FC triferroics, with moderate energy barriers of 0.129 and 0.363 eV atom-1 for FE and FC switching, respectively. Both FE and FC switching can reverse the AM spin splitting in antiferromagnetic (AFM) configurations while preserving the high spin polarization in FM states. The magnetic anisotropy is highly tunable, exhibiting either uniaxial or in-plane behavior depending on the phase, dimension, and facet. This work establishes a framework for designing AM multiferroics through polymorphic, dimensional, and facet engineering, offering promising avenues for multifunctional spintronic applications.

cond-mat.mtrl-sci↗

arXiv2025

STEP: Success-Rate-Aware Trajectory-Efficient Policy Optimization

Multi-turn interaction remains challenging for online reinforcement learning. A common solution is trajectory-level optimization, which treats each trajectory as a single training sample. However, this approach can be inefficient and yield misleading learning signals: it applies uniform sampling across tasks regardless of difficulty, penalizes correct intermediate actions in failed trajectories, and incurs high sample-collection costs. To address these issues, we propose STEP (Success-rate-aware Trajectory-Efficient Policy optimization), a framework that dynamically allocates sampling based on per-task success rates and performs step-level optimization. STEP maintains a smoothed success-rate record to guide adaptive trajectory resampling, allocating more effort to harder tasks. It then computes success-rate-weighted advantages and decomposes trajectories into step-level samples. Finally, it applies a step-level GRPO augmentation to refine updates for low-success tasks. Experiments on OSWorld and AndroidWorld show that STEP substantially improves sample efficiency and training stability over trajectory-level GRPO, converging faster and generalizing better under the same sampling budget.

cs.AI↗

arXiv2025

Universal Behavior in Entanglement Entropy Reveals Quantum Criticality and Underlying Symmetry Breaking

Entanglement plays a key role in quantum physics, but how much information it can extract from many-body systems is still an open question, particularly regarding quantum criticalities and emergent symmetries. In this work, we systematically study the entanglement entropy (EE) and derivative entanglement entropy (DEE) near quantum phase transitions in various quantum many-body systems. A one-parameter scaling relation between the DEE and system size at the critical point has been derived for the first time, which successfully obtains the critical exponent via data collapse. Furthermore, we find that the EE peaks at the (emergent) symmetry enhanced first-order transition, reflecting higher symmetry breaking. This work provides a new paradigm for quantum many-body research from the perspective of EE and DEE.

cond-mat.str-el↗

arXiv2025

Lithium and Vanadium Intercalation into Bilayer V2Se2O: Ferrimagnetic-Ferroelastic Multiferroics and Anomalous and Spin Transport

Spin splitting in emerging altermagnets is non-relativistic and momentum-dependent, yet energy-independent, and localized in momentum space, posing challenges for practical applications. Here, we propose an intercalation-driven paradigm for altermagnets to attain ameliorative electronic structures, multiferroic characteristics, and anomalous and spin transport functionalities. As a representative system, we investigate electrochemistry- and self-intercalated V2Se2O bilayers, building on the recently reported room-temperature K- and Rb-intercalated V2Se2O family [Nat. Phys. 2025, 21, 754; Nat. Phys. 2025, 21, 760], utilizing density functional theory, Wannier function analyses, Monte Carlo simulations, and non-equilibrium Green function methods. Intercalation induces room-temperature intralayer ferrimagnetic and interlayer ferromagnetic order (358 K for Li-intercalation and 773 K for V-intercalation), ferroelasticity (~1 % signal intensity), in-plane uniaxial magnetic anisotropy, and metallization, while also modifying the anomalous Hall effect. Notably, Li- and V-intercalated V2Se2O bilayers exhibit enhanced spin splitting and half-metallic behavior, respectively, yielding near-perfect spin filtering efficiency. Intercalation substantially enhances spin transport in V2Se2O-based devices, enabling giant magnetoresistance (877 %), ultra-high thermal tunneling magnetoresistance (~12000 %), and observable spin Seebeck and temperature negative differential resistance effects. This intercalation-driven paradigm expands altermagnetic functionalities through multifunctional integration, offering promising avenues for advanced, miniaturized, room-temperature exploitation of anomalous, electron, and spin transport properties.

cond-mat.mtrl-sci↗

arXiv2025

Above room temperature multiferroic tunnel junction with the altermagnetic metal CrSb

Altermagnets with nonrelativistic momentum-dependent spin splitting and compensated net magnetic moments have recently garnered significant interest in spintronics, particularly as pinning layers in magnetic tunnel junctions (MTJs). However, room temperature (RT) altermagnet-based MTJs with tunable tunneling magnetoresistance (TMR) or electroresistance (TER) modulated by multiferroicity remain largely unexplored. Here, we propose an experimentally fabricable above-RT multiferroic MTJ, comprising an altermagnetic metal, ferroelectric barrier, and ferromagnetic metal-epitomized by a CrSb/In2Se3/Fe3GaTe2 heterostructure. Our calculations with first-principles and nonequilibrium Green function method indicate that the architecture enables magnetically switchable TER, electrically tunable TMR, and dual-mode controllable spin filtering. To disentangle the roles of ferroelectricity and the tunnel barrier, nonferroelectric Sb2Se3 and a vacuum gap are exploited as control cases. Remarkably, the system achieves TMR up to 2308%, TER of 707%, and near-perfect spin filtering efficiency. Both TMR and TER are considerable for CrSb/In2Se3/Fe3GaTe2 with either Cr or Sb interface. The transport performance is robust under bias voltage. These findings demonstrate the above-RT multiferroic altermagnet-based MTJs and highlight their exciting potential as a versatile platform for next-generation spin dynamics, magnetic sensing, and quantum logic nanodevices.

cond-mat.mtrl-sci↗

arXiv2025

Context-Aware Pseudo-Label Scoring for Zero-Shot Video Summarization

We propose a rubric-guided, pseudo-labeled, and prompt-driven zero-shot video summarization framework that bridges large language models with structured semantic reasoning. A small subset of human annotations is converted into high-confidence pseudo labels and organized into dataset-adaptive rubrics defining clear evaluation dimensions such as thematic relevance, action detail, and narrative progression. During inference, boundary scenes, including the opening and closing segments, are scored independently based on their own descriptions, while intermediate scenes incorporate concise summaries of adjacent segments to assess narrative continuity and redundancy. This design enables the language model to balance local salience with global coherence without any parameter tuning. Across three benchmarks, the proposed method achieves stable and competitive results, with F1 scores of 57.58 on SumMe, 63.05 on TVSum, and 53.79 on QFVS, surpassing zero-shot baselines by +0.85, +0.84, and +0.37, respectively. These outcomes demonstrate that rubric-guided pseudo labeling combined with contextual prompting effectively stabilizes LLM-based scoring and establishes a general, interpretable, and training-free paradigm for both generic and query-focused video summarization.

cs.CV↗

arXiv2025

Spinons, solitons and random singlets in the spin-chain compound copper benzoate

The $S=1/2$ antiferromagnetic Heisenberg chain is a paradigmatic quantum system hosting exotic excitations such as spinons and solitons, and forming random singlet state in the presence of quenched disorder. Realizing and distinguishing these excitations in a single material remains a significant challenge. Using nuclear magnetic resonance (NMR) on a high-quality single crystal of copper benzoate, we identify and characterize all three excitation types by tuning the magnetic field at ultra-low temperatures. At a low field of 0.2 T, a temperature-independent spin-lattice relaxation rate ($1/T_1$) over more than a decade confirms the presence of spinons. Below 0.4 K, an additional relaxation channel emerges, characterized by $1/T_1 \propto T$ and a spectral weight growing as $-\ln(T/T_0)$, signaling a random-singlet ground state induced by weak quenched disorder. At fields above 0.5 T, a field-induced spin gap $Δ\propto H^{2/3}$ observed in both $1/T_1$ and the Knight shift signifies soliton excitations. Our results establish copper benzoate as a unique experimental platform for studying one-dimensional quantum integrability and the interplay of disorder and correlations.

cond-mat.str-el↗

arXiv2025

CPCLDETECTOR: Knowledge Enhancement and Alignment Selection for Chinese Patronizing and Condescending Language Detection

Chinese Patronizing and Condescending Language (CPCL) is an implicitly discriminatory toxic speech targeting vulnerable groups on Chinese video platforms. The existing dataset lacks user comments, which are a direct reflection of video content. This undermines the model's understanding of video content and results in the failure to detect some CPLC videos. To make up for this loss, this research reconstructs a new dataset PCLMMPLUS that includes 103k comment entries and expands the dataset size. We also propose the CPCLDetector model with alignment selection and knowledge-enhanced comment content modules. Extensive experiments show the proposed CPCLDetector outperforms the SOTA on PCLMM and achieves higher performance on PCLMMPLUS . CPLC videos are detected more accurately, supporting content governance and protecting vulnerable groups. Code and dataset are available at https://github.com/jiaxunyang256/PCLD.

cs.MM↗

arXiv2025

TsqLoRA: Towards Sensitivity and Quality Low-Rank Adaptation for Efficient Fine-Tuning

Fine-tuning large pre-trained models for downstream tasks has become a fundamental approach in natural language processing. Fully fine-tuning all model parameters is computationally expensive and memory-intensive, especially in resource-constrained environments. Existing parameter-efficient fine-tuning methods reduce the number of trainable parameters but typically overlook the varying sensitivity of different model layers and the importance of training data. In this work, we propose TsqLoRA, a novel method that integrates data-quality-driven selection with sensitivity-aware low-rank adaptation, consisted of two main components: a quality-aware sampling mechanism for selecting the most informative training data, and a dynamic rank allocation module that adjusts the rank of each layer based on its sensitivity to parameter updates. The experimental results demonstrate that TsqLoRA improves fine-tuning efficiency while maintaining or even improving performance on a variety of NLP tasks. Our code will be available at https://github.com/Benjamin-Ricky/TsqLoRA.

cs.CL↗

arXiv2025

LLM-OREF: An Open Relation Extraction Framework Based on Large Language Models

The goal of open relation extraction (OpenRE) is to develop an RE model that can generalize to new relations not encountered during training. Existing studies primarily formulate OpenRE as a clustering task. They first cluster all test instances based on the similarity between the instances, and then manually assign a new relation to each cluster. However, their reliance on human annotation limits their practicality. In this paper, we propose an OpenRE framework based on large language models (LLMs), which directly predicts new relations for test instances by leveraging their strong language understanding and generation abilities, without human intervention. Specifically, our framework consists of two core components: (1) a relation discoverer (RD), designed to predict new relations for test instances based on \textit{demonstrations} formed by training instances with known relations; and (2) a relation predictor (RP), used to select the most likely relation for a test instance from $n$ candidate relations, guided by \textit{demonstrations} composed of their instances. To enhance the ability of our framework to predict new relations, we design a self-correcting inference strategy composed of three stages: relation discovery, relation denoising, and relation prediction. In the first stage, we use RD to preliminarily predict new relations for all test instances. Next, we apply RP to select some high-reliability test instances for each new relation from the prediction results of RD through a cross-validation method. During the third stage, we employ RP to re-predict the relations of all test instances based on the demonstrations constructed from these reliable test instances. Extensive experiments on three OpenRE datasets demonstrate the effectiveness of our framework. We release our code at https://github.com/XMUDeepLIT/LLM-OREF.git.

cs.CL↗

arXiv2025

10-W Sub-100-fs Ultrafast Cr:ZnS/ZnSe MOPA System enabled by doping gradient engineering

We report on a high-power mid-infrared femtosecond master oscillator power amplifier (MOPA) system, employing Cr:ZnS and Cr:ZnSe polycrystals with fine-tuned doping profiles. Based on the soft-aperture Kerr-lens mode-locking in the soliton regime, the seed oscillator generates ~40-fs pulses with a repetition rate ~173 MHz with an average power close to 400 mW. The amplification process of the seed pulse train is investigated in depth in a single-pass configuration for both Cr:ZnS and Cr:ZnSe crystal rods. For further power scaling, a dual-stage MOPA system has been implemented, generating pulse trains with an average power up to 10.4 W, limited only by the pump source, with a re-compressed pulse duration of 78 fs using a dispersion compensator comprising chirped mirrors and sapphire plates. This work paves the way for further power scaling of mid-infrared Cr:ZnS/ZnSe ultrafast laser systems without moving parts for applications in material processing, remote sensing and medicine.

physics.optics↗

arXiv2025

GeoGPT-RAG Technical Report

GeoGPT is an open large language model system built to advance research in the geosciences. To enhance its domain-specific capabilities, we integrated Retrieval Augmented Generation(RAG), which augments model outputs with relevant information retrieved from an external knowledge source. GeoGPT uses RAG to draw from the GeoGPT Library, a specialized corpus curated for geoscientific content, enabling it to generate accurate, context-specific answers. Users can also create personalized knowledge bases by uploading their own publication lists, allowing GeoGPT to retrieve and respond using user-provided materials. To further improve retrieval quality and domain alignment, we fine-tuned both the embedding model and a ranking model that scores retrieved passages by relevance to the query. These enhancements optimize RAG for geoscience applications and significantly improve the system's ability to deliver precise and trustworthy outputs. GeoGPT reflects a strong commitment to open science through its emphasis on collaboration, transparency, and community driven development. As part of this commitment, we have open-sourced two core RAG components-GeoEmbedding and GeoReranker-to support geoscientists, researchers, and professionals worldwide with powerful, accessible AI tools.

cs.IR↗

arXiv2025

Optical Fiber-Waveguide Hybrid Architecture for Mid-Infrared Ring Lasers

We report a mid-infrared ring laser via fiber-to-waveguide hybrid integration enabled by directional couplers fabricated in fluoride glasses. Directional couplers were inscribed into fluoride glass using femtosecond laser direct writing, forming low-loss waveguides with a high positive refractive index change. The waveguides were optimized for efficient guidance at a wavelength of 2.7 {\textmu}m. Directional couplers with systematically varied coupling gap and interaction lengths were fabricated and experimentally characterized, demonstrating controllable power splitting ratios spanning from complete transmission (100:0) to balanced coupling (50:50). A directional coupler with a 7 {\textmu}m coupling gap and a 1.5 mm interaction length, providing an 80:20 power splitting ratio, was integrated with a cladding-pumped Er:ZBLAN fiber to realize a ring laser cavity. The resulting device produced stable lasing at 2716.7 nm with an output power of 14.2 mW. This work represents the first realization of a fully fiber-chip integrated mid-infrared laser source and opens new avenues for 3D photonic integration in fluoride glasses, enabling design flexibility previously unattainable in traditional fiber-optics architectures.

physics.optics↗

arXiv2025

Measuring topological invariants of even-dimensional line-gapped non-Hermitian systems through quench dynamics

The accurate determination of non-Hermitian (NH) topological invariants plays a central role in the study of NH topological phases. In this work, we propose a general framework for directly measuring NH topological invariants in even-dimensional systems with real line gaps through quench dynamics. Our approach hinges on constructing an auxiliary Hermitian matrix topologically equivalent to the original NH Hamiltonian, enabling topological characterization via reduced-dimensional momentum subspaces called band-inversion surfaces (BISs). A key insight lies in the emergence of chiral symmetry in the NH Hamiltonian specifically on BISs -- a critical property that allows extension of the dynamical characterization scheme previously developed for odd-dimensional NH systems with chiral or sublattice symmetry [Lin et al., Phys. Rev. Res. 7, L012060 (2025)]. We show that NH topological invariants can be extracted from the winding patterns of a dynamical field constructed from post-quench spin textures on BISs. We demonstrate our approach through a detailed analysis of NH Chern insulators and then extend the framework to higher even-dimensional systems by introducing second-order BISs for characterization. The framework is also generalized to imaginary line-gapped topological phases. This work establishes an experimentally accessible protocol for detecting NH topological invariants in quantum platforms.

cond-mat.quant-gas↗

arXiv2025

Enhancing Partially Relevant Video Retrieval with Robust Alignment Learning

Partially Relevant Video Retrieval (PRVR) aims to retrieve untrimmed videos partially relevant to a given query. The core challenge lies in learning robust query-video alignment against spurious semantic correlations arising from inherent data uncertainty: 1) query ambiguity, where the query incompletely characterizes the target video and often contains uninformative tokens, and 2) partial video relevance, where abundant query-irrelevant segments introduce contextual noise in cross-modal alignment. Existing methods often focus on enhancing multi-scale clip representations and retrieving the most relevant clip. However, the inherent data uncertainty in PRVR renders them vulnerable to distractor videos with spurious similarities, leading to suboptimal performance. To fill this research gap, we propose Robust Alignment Learning (RAL) framework, which explicitly models the uncertainty in data. Key innovations include: 1) we pioneer probabilistic modeling for PRVR by encoding videos and queries as multivariate Gaussian distributions. This not only quantifies data uncertainty but also enables proxy-level matching to capture the variability in cross-modal correspondences; 2) we consider the heterogeneous informativeness of query words and introduce learnable confidence gates to dynamically weight similarity. As a plug-and-play solution, RAL can be seamlessly integrated into the existing architectures. Extensive experiments across diverse retrieval backbones demonstrate its effectiveness.

cs.CV↗

arXiv2025

WideSearch: Benchmarking Agentic Broad Info-Seeking

From professional research to everyday planning, many tasks are bottlenecked by wide-scale information seeking, which is more repetitive than cognitively complex. With the rapid development of Large Language Models (LLMs), automated search agents powered by LLMs offer a promising solution to liberate humans from this tedious work. However, the capability of these agents to perform such "wide-context" collection reliably and completely remains largely unevaluated due to a lack of suitable benchmarks. To bridge this gap, we introduce WideSearch, a new benchmark engineered to evaluate agent reliability on these large-scale collection tasks. The benchmark features 200 manually curated questions (100 in English, 100 in Chinese) from over 15 diverse domains, grounded in real user queries. Each task requires agents to collect large-scale atomic information, which could be verified one by one objectively, and arrange it into a well-organized output. A rigorous five-stage quality control pipeline ensures the difficulty, completeness, and verifiability of the dataset. We benchmark over 10 state-of-the-art agentic search systems, including single-agent, multi-agent frameworks, and end-to-end commercial systems. Most systems achieve overall success rates near 0\%, with the best performer reaching just 5\%. However, given sufficient time, cross-validation by multiple human testers can achieve a near 100\% success rate. These results demonstrate that present search agents have critical deficiencies in large-scale information seeking, underscoring urgent areas for future research and development in agentic search. Our dataset, evaluation pipeline, and benchmark results have been publicly released at https://widesearch-seed.github.io/

cs.CL↗

arXiv2025

Compass-Thinker-7B Technical Report

Recent R1-Zero-like research further demonstrates that reasoning extension has given large language models (LLMs) unprecedented reasoning capabilities, and Reinforcement Learning is the core technology to elicit its complex reasoning. However, conducting RL experiments directly on hyperscale models involves high computational costs and resource demands, posing significant risks. We propose the Compass-Thinker-7B model, which aims to explore the potential of Reinforcement Learning with less computational resources and costs, and provides insights for further research into RL recipes for larger models. Compass-Thinker-7B is trained from an open source model through a specially designed Reinforcement Learning Pipeline. We curate a dataset of 30k verifiable mathematics problems for the Reinforcement Learning Pipeline. By configuring data and training settings with different difficulty distributions for different stages, the potential of the model is gradually released and the training efficiency is improved. Extensive evaluations show that Compass-Thinker-7B possesses exceptional reasoning potential, and achieves superior performance on mathematics compared to the same-sized RL model. Especially in the challenging AIME2024 evaluation, Compass-Thinker-7B achieves 40% accuracy.

cs.AI↗