arXiv subjects

Qi Zhang

Publications and source records attributed to Qi Zhang.

At least 127 records · Page 7Linked to original sources

arXiv2026

Adaptive Visual Autoregressive Acceleration via Dual-Linkage Entropy Analysis

Visual AutoRegressive modeling (VAR) suffers from substantial computational cost due to the massive token count involved. Failing to account for the continuous evolution of modeling dynamics, existing VAR token reduction methods face three key limitations: heuristic stage partition, non-adaptive schedules, and limited acceleration scope, thereby leaving significant acceleration potential untapped. Since entropy variation intrinsically reflects the transition of predictive uncertainty, it offers a principled measure to capture modeling dynamics evolution. Therefore, we propose NOVA, a training-free token reduction acceleration framework for VAR models via entropy analysis. NOVA adaptively determines the acceleration activation scale during inference by online identifying the inflection point of scale entropy growth. Through scale-linkage and layer-linkage ratio adjustment, NOVA dynamically computes distinct token reduction ratios for each scale and layer, pruning low-entropy tokens while reusing the cache derived from the residuals at the prior scale to accelerate inference and maintain generation quality. Extensive experiments and analyses validate NOVA as a simple yet effective training-free acceleration framework.

cs.CV↗

arXiv2026

Robust logical Bell nonlocality based on quantum error correction codes

Quantum nonlocality based on the violation of Bell-like inequalities constitutes a fundamental feature of quantum physics and drives the development of device-independent (DI) quantum information technologies. Existing studies of Bell nonlocality have mainly focused on physical qubit systems, where the observed nonlocal correlations are directly encoded in physical degrees of freedom. The decoherence sensitivity of Bell nonlocality largely limits the performance and security of its DI applications. Here, we investigate the robust logical Bell nonlocality based on quantum error correction codes. We construct the general logical Bell inequality in the stabilizer coding subspace and prove its violation indicates the global nonlocal feature of the logical system. Then, we indicate that the logical Bell nonlocality is robust against decoherence. Comparing with the physical qubit system, the fidelity thresholds for the logical Bell inequality violation based on the [[3,1,1]] and [[7,1,1]] repetition codes under the bit-flip error model can be reduced from 82.8% to 73.10% and 66.35%, increasing DI QKD's bit-flip noise threshold from 10.64% to 14.42% and 23.36%, respectively. Such stabilizer-based framework can be also used to characterize the multipartite logical Bell nonlocality in principle. Finally, a logical Bell test implementation circuit based on the [[3,1,1]] repetition code is presented. This work provides a feasible avenue for unlocking robust Bell nonlocality in scalable logical quantum systems and facilitates its applications in future scalable quantum network.

quant-ph↗

arXiv2026

AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilities

As Large Language Models (LLMs) evolve into autonomous agents, the need for unified evaluation infrastructure becomes critical. However, current evaluation pipelines remain highly fragmented and tightly coupled, hindering reproducibility and causing redundant engineering. To address this, we introduce AgentCompass, an open-source, lightweight, and extensible infrastructure for evaluating LLM-based agents. AgentCompass organizes the evaluation process around three independent components, namely Benchmark, Harness, and Environment, thereby enabling flexible configurations without requiring the reimplementation of complex execution logic. Furthermore, it features a fault-tolerant asynchronous runtime and comprehensive trajectory analysis tools to transparently diagnose nuanced failure modes like reward-hacking. Natively supporting over 20 benchmarks across five capability dimensions, AgentCompass provides the community with a scalable and reproducible infrastructure for advancing agent research.

cs.AI↗

arXiv2026

Observation of cooperative strong coupling between optical phonon and crystal-field excitations in a pseudo Jahn-Teller system

Cooperative interactions between localized electronic excitations and crystal lattice are central to the emergence of complex structural phases in materials. However, the scaling relations governing these collective behaviors remain largely unexplored. Here, using magneto-Raman spectroscopy, we report the direct observation of the strongly coupled optical phonon and non-degenerate crystal-field excitations (CFEs) in ErFeO3. By independently tuning the effective population of Jahn-Teller-active erbium ions through temperature and chemical dilution with Jahn-Teller-inactive yttrium ions, we identify the coupling strength varies linearly with the square root of electronic excitations population. Notably, Y-doping reveals the hybridization gap reduces significantly faster than predicted by density scaling alone, indicating phonon coherence is essential for establishing this cooperative interaction. Our findings highlight the role of optical phonons in mediating short-range interactions that drive cooperative Jahn-Teller effect, evidencing the pathway for tailoring electronic and vibrational properties of Jahn-Teller materials through population control.

cond-mat.mtrl-sci↗

arXiv2026

Modality Discrepancy Transformer for Ambivalence and Hesitancy Recognition

Ambivalence and hesitancy (A/H) are affective states in which individuals express contradictory signals across facial, vocal, and linguistic channels. Automatically recognising A/H in clinical videos requires detecting cross-modal disagreement -- the signal that standard fusion methods suppress. Based on the conflict-aware multimodal fusion framework of Bekhouche et al., we present the Modality Discrepancy Transformer (MDT). MDT enriches the original 6-token design to a 9-token representation comprising three modality embeddings, three absolute-difference features, and three Hadamard-product discrepancy features learned through linear projections. These nine tokens undergo Transformer self-attention, with FiLM-based text-conditioned modulation and LoRA fine-tuning as core architectural components. A text-guided late fusion branch blends a text-only auxiliary head with the full multimodal output at inference. On the BAH dataset from the 3rd ABAW Challenge, MDT achieves 0.7408 Macro F1 on the labelled test split and 0.7368 on the private leaderboard, outperforming the strongest published baseline by over 10 points while training in under 20 minutes on a single GPU.

cs.CL↗

arXiv2026

CoWeaver: A Bi-directional, Learnable and Explainable Matching Engine for Mixed Human-Agent Science Collaboration

LLM-based agents excel at writing articles, coding and information retrieval. However, they fail to form strong collaborations within the scientific community due to the bidirectional, dynamic nature of the problem and a high demand of decision interpretability. We proposed COWEAVER, a bidirectional, learnable and explainable algorithm to match scientists and form strong collaborations within a human-agent network. COWEAVER matches candidates and requesters through filling capability gaps and filters candidates through a two-stage ranking step. Finally, the model explores newcomers by maintaining uncertainty-aware capability estimates and updating them through requester's feedback. We show that the selection mechanism of combining both exploration (UCB) and greedy of COWEAVER exceeds the greedy-only mechanism - the analytical best solution - on 6 out of the 20 tasks and performed on par with the greedy-only mechanism in terms of selecting the best candidate. We compared COWEAVER baselines in terms of matching quality and efficiency. COWEAVER outperforms baselines on all metrics.

cs.MA↗

arXiv2026

HELP: Human-Efficient Large-Scale Robot Post-Training with Rollout Segmentation

When adapting Vision Language Action (VLA) models to downstream tasks, multiple rounds of post-training are often required to progressively address policy weaknesses. In this report, we focus on maximizing human efficiency during this iterative process, measured by policy improvement and task throughput per unit of human labor and time. We propose HELP, a Human-Efficient Large-scale robot Post-training pipeline in which two specialized operators supervise twelve robots concurrently. A trained Teleoperator provides high-value remote interventions and recovery demonstrations, while a Floor Operator monitors the robot fleet, triggers takeovers, and performs physical resets. This role specialization improves human efficiency by reducing task switching, lowering operator training costs, and expanding robot interaction coverage. Beyond increasing rollout volume, concurrent supervision also broadens the range of policy behaviors observed by the human team, making recurring failure modes easier to identify and enabling more targeted takeovers, resets, and recovery demonstrations. To efficiently utilize the large and mixed-quality rollout data, HELP incorporates \vlac, an automatic rollout segmentation critic specifically designed for this setting. It separates autonomous trajectories into progress-making, idle, failure-inducing, and recovery segments. Useful rollout segments are retained and combined with Human-in-the-Loop data for the next post-training round. Across four real-world manipulation tasks, HELP achieves 80\%--95\% success rates and improves task throughput by 1.7$\times$--4.2$\times$ over the base model. Under matched HITL recovery budgets, VLAC-CUT further amplifies throughput gains by 1.20$\times$--3.43$\times$ and success-rate gains by 1.50$\times$--3.00$\times$ over HITL-only updates.

cs.RO↗

arXiv2026

Ferroelectric--Superconducting Interaction in Epitaxial YBa2Cu3O7-δ/BaTiO3 Films

Microscopic theories predict that the critical temperature of a superconducting layer can be strongly modified in proximity to a ferroelectric; however, experimental evidence is lacking. We report on BaTiO3 (BTO)/YBa2Cu3O7-δ (YBCO) heterostructures grown on SrTiO3 (001) substrates using pulsed laser deposition (PLD), with precise control of growth conditions and oxygen stoichiometry. Piezoresponse force microscopy confirms ferroelectric switching within both single-layer and heterostructured films. Neutron reflectometry measurements show low roughness parameters for the buried interfaces, which is critical for mediating the interfacial coupling between superconductivity and ferroelectric polarization. The superconducting transition temperature ($T_{\mathrm{C}}^{R}$) measured from the $R(T)$ curve of the single-layer YBCO was approximately 86 K, and a transition temperature ($T_{\mathrm{C}}^{M}$) of approximately 80-85 K was observed from the $M(T)$ curve. For the heterostructure film (YBCO/BTO/YBCO) grown under the same conditions, both resistivity and magnetization measurements indicate lower transition temperatures of approximately 75 K and approximately 55 K, respectively, as well as a lower overall susceptibility. These findings highlight the need to combine multiple techniques to characterize BTO/YBCO heterostructures and identify subtle features of the ferroelectric--superconducting coupling. This establishes BTO/YBCO heterostructures as promising systems for tunable oxide electronics and reconfigurable superconducting devices.

cond-mat.supr-con↗

arXiv2026

Super-Generalist: Towards Comprehensive and Accurate Medical Image Understanding via Generalist-Specialist Synergy

Medical images require comprehensive and accurate interpretation to support the diagnosis of diverse clincial conditions. Recent vision-language generalist models offer broad task coverage and promising zero-shot capabilities, yet often lack fine-grained anatomical and lesion awareness for reliable diagnosis and spatial interpretability. In contrast, supervised specialist models achieve strong performance on specific tasks but typically lack generalization across diseases and anatomies. In this work, we present SuG, a Super-Generalist framework that unifies generalist vision-language learning with specialist objectives, enabling both broad generalization and specialist-level diagnostic capability. We perform specialist-enhanced vision-language alignment in SuG by incorporating spatial priors from multiple segmentation experts, including anatomy, class-specific lesion and class-agnostic lesion segmentors that captures lesions beyond anatomies annotated during training. To improve lesion grounding capability, we leverage lesion masks as spatial priors to calibrate text-conditioned visual attention, encouraging disease-related semantics to focus on clinically relevant regions. We evaluate SuG on extensive chest and abdominal CT benchmarks, including CT-RATE, Merlin, MedVL-CT69K, and several in-house tumor datasets. SuG achieves state-of-the-art performance across a wide range of disease diagnosis tasks and surpasses specialist models on several critical tumor diagnosis benchmarks. Furthermore, SuG demonstrates strong lesion grounding capability, including robust generalization to lesion types lacking class-specific supervision.

cs.CV↗

arXiv2026

Bridging Cognitive Neuroscience and Graph Intelligence: Hippocampus-Inspired Multi-View Hypergraph Learning for Web Finance Fraud

Online financial services constitute an essential component of contemporary web ecosystems, yet their openness introduces substantial exposure to fraud that harms vulnerable users and weakens trust in digital finance. Such threats have become a significant web harm that erodes societal fairness and affects the well-being of online communities. However, existing detection methods based on graph neural networks (GNNs) struggle with two persistent challenges: (1) long-tailed data distributions, which obscure rare but critical fraudulent cases, and (2) fraud camouflage, where malicious transactions mimic benign behaviors to evade detection. To fill these gaps, we propose HIMVH, a Hippocampus-Inspired Multi-View Hypergraph learning model for web finance fraud detection. Specifically, drawing inspiration from the scene conflict monitoring role of the hippocampus, we design a cross-view inconsistency perception module that captures subtle discrepancies and behavioral heterogeneity across multiple transaction views. This module enables the model to identify subtle cross-view conflicts for detecting online camouflaged fraudulent behaviors. Furthermore, inspired by the match-mismatch novelty detection mechanism of the CA1 region, we introduce a novelty-aware hypergraph learning module that measures feature deviations from neighborhood expectations and adaptively reweights messages, thereby enhancing sensitivity to online rare fraud patterns in the long-tailed settings. Extensive experiments on six web-based financial fraud datasets demonstrate that HIMVH achieves 6.42% improvement in AUC, 9.74% in F1 and 39.14% in AP on average over 15 SOTA models.

cs.LG↗

arXiv2026

Efficient photo-ionizing elimination of detrimental electric fields for Rydberg atoms

Rydberg atoms are highly sensitive to external electric fields due to their exaggerated electronic properties. This unique feature lays the foundation for many of their applications in quantum science. However, an uncontrolled stray electric field can be detrimental, severely degrading their quantum control. In this work, we demonstrate a universal scheme that relies on the efficient creation of an in-vacuum plasma source by photo-ionizing laser-cooled atoms to eliminate detrimental electric fields in a Rydberg-atom tweezer array platform, requiring only readily available resources. With this method, we began with a Stark-ionized Rydberg continuum spectrum caused by a large, unknown stray electric field and ultimately recovered stable, coherent excitation of an individual Rydberg state after fully eliminating the field. Our method is directly applicable to existing Rydberg-atom platforms and can also be useful in other experiments sensitive to stray electric fields.

physics.atom-ph↗

arXiv2026

RIMRULE: Improving Tool-Using Language Agents via MDL-Guided Rule Learning

Large language models (LLMs) often struggle to use tools reliably in domain-specific settings, where APIs may be idiosyncratic, under-documented, or tailored to private workflows. This highlights the need for effective adaptation to task-specific tools. We propose RIMRULE, a neuro-symbolic approach for LLM adaptation based on dynamic rule injection. Compact, interpretable rules are distilled from failure traces and injected into the prompt during inference to improve task performance. These rules are proposed by the LLM itself and consolidated using a Minimum Description Length (MDL) objective that favors generality and conciseness. Each rule is stored in both natural language and a structured symbolic form, supporting efficient retrieval at inference time. Experiments on tool-use benchmarks show that this approach improves accuracy on both seen and unseen tools without modifying LLM weights. It outperforms prompting-based adaptation methods and complements finetuning. Moreover, rules learned from one LLM can be reused to improve others, including long reasoning LLMs, highlighting the portability of symbolic knowledge across architectures.

cs.CL↗

arXiv2026

AgentGym2: Benchmarking Large Language Model Agents in De-Idealized Real-World Environments

Language agents, i.e., LLM agents, progress rapidly and are increasingly deployed in production environments. This trend underscores the urgent need for rigorous and realistic evaluations. However, most existing benchmarks evaluate agents in simplified, idealized settings. They typically rely on pre-packaged tool interfaces, overlook critical steps, and assume inputs are clean and fully specified. Consequently, they understate the difficulty of real deployments, where uncertainty and noise are ubiquitous and agents must proactively explore the environment to uncover new tools. To bridge this gap, we present AgentGym2, a new evaluation framework with task instances grounded in real-world end-to-end working demands. Beyond reasoning and planning, it measures agents' ability to execute end-to-end procedures, discover tools via exploration, compose tools for unseen tasks, and remain robust to noisy and underspecified information. Experiments on 15 proprietary and open-source models show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2, revealing a substantial gap between the capability of current agents and the demands of real-world applications.

cs.AI↗

arXiv2026

Kwai Summary Attention Technical Report

Long-context ability, has become one of the most important iteration direction of next-generation Large Language Models, particularly in semantic understanding/reasoning, code agentic intelligence and recommendation system. However, the standard softmax attention exhibits quadratic time complexity with respect to sequence length. As the sequence length increases, this incurs substantial overhead in long-context settings, leading the training and inference costs of extremely long sequences deteriorate rapidly. Existing solutions mitigate this issue through two technique routings: i) Reducing the KV cache per layer, such as from the head-level compression GQA, and the embedding dimension-level compression MLA, but the KV cache remains linearly dependent on the sequence length at a 1:1 ratio. ii) Interleaving with KV Cache friendly architecture, such as local attention SWA, linear kernel GDN, but often involve trade-offs among KV Cache and long-context modeling effectiveness. Besides the two technique routings, we argue that there exists an intermediate path not well explored: {Maintaining a linear relationship between the KV cache and sequence length, but performing semantic-level compression through a specific ratio $k$}. This $O(n/k)$ path does not pursue a ``minimum KV cache'', but rather trades acceptable memory costs for complete, referential, and interpretable retention of long distant dependency. Motivated by this, we propose Kwai Summary Attention (KSA), a novel attention mechanism that reduces sequence modeling cost by compressing historical contexts into learnable summary tokens.

cs.CL↗

arXiv2026

Electron-beam Writing of Spectrally Uniform Green Single-photon Emitters in Hexagonal Boron Nitride

Scalable quantum photonic technologies require single-photon emitters whose positions and emission energies can be engineered simultaneously. Hexagonal boron nitride (hBN) is an attractive room-temperature host, but deterministic creation of spectrally reproducible emitters remains challenging. Here, we use a standard scanning electron microscope as a direct-writing tool to activate bright green single-photon emitters in hBN at predefined sites, without ion implantation or post-fabrication thermal annealing. The written emitters exhibit reproducible zero-phonon-line emission centered near 536 nm, room-temperature antibunching with g(2)(0) as low as 0.08, high brightness, strong linear polarization, and stable emission. Thickness-dependent activation, stacking experiments, cathodoluminescence spectroscopy, and first-principles calculations support a carbon-related defect complex as the most plausible origin of the emission. As a proof of nanophotonic compatibility, we further activate emitters in a nanoparticle-on-mirror plasmonic nanocavity and observe photoluminescence enhancement accompanied by shortened emission lifetimes. These results establish electron-beam direct writing as a practical route to site-selective, spectrally uniform green quantum emitters in hBN, offering a promising basis for integrated room-temperature quantum photonic architectures.

physics.optics↗

arXiv2026

Free energy dissipation and a decomposition of general jump diffusions on $\mathbb{R}^n$ without detailed balance

We analyze the thermodynamic structure of jump diffusions combining Brownian and Poisson noise, a class of stochastic dynamics relevant to non-equilibrium statistical physics. For such nonlocal dynamics, the free energy admits a full dissipation formula that decomposes into entropy production and housekeeping heat. A central result is a decomposition of the generator into symmetric and anti-symmetric parts with respect to the invariant measure $ρ_\mathrm{ss}$. The symmetric sector corresponds to a reversible dynamics and yields a nonlocal Fisher information governing free-energy decay, whereas the anti-symmetric sector generates a canonical conservative flow that produces circulation but no dissipation. Several numerical examples motivated by intracellular particle transports demonstrate how this decomposition clarifies the structure of non-equilibrium stationary states in jump-driven systems.

cond-mat.stat-mech↗

arXiv2026

Neural Canonical Transformation for the Spectra of Fluxional Molecule CH5+

Protonated methane, CH5+, is a highly fluxional molecule with large spatial motions of the hydrogen atoms. The molecule's anharmonic effects and the delocalized wavefunction of the hydrogen atoms significantly affect the excitation spectrum of the molecule. The neural canonical transformation (NCT) approach, which we previously developed to solve the vibrational spectra of molecules and solids, is a powerful method that effectively treats nuclear quantum effects and anharmonicities. Using NCT with wavefunctions in atomic coordinates rather than normal coordinates, we successfully calculate the ground and excited states of CH5+. We found that the wavefunctions for the ground state, as well as for low- and high-energy excited states, show preferences for the three stationary points on the potential energy surface. This work extends the applicability of the NCT approach for calculating excited states to fluxional molecules without fixed geometry.

physics.chem-ph↗

arXiv2026

Quantum LiDAR with non-local modulation

Quantum light detection and ranging (LiDAR) utilizes quantum entanglement and correlation to improve precision, noise resilience and covertness of target detection. Despite recent advances, the development of a quantum LiDAR system that simultaneously achieves high precision and a large measurement range remains challenging. Here, we demonstrate a quantum amplitude-modulated continuous wave LiDAR with micrometer precision achievable via increased acquisition time and meter-scale measurement range. In our demonstration, the signal photons directly illuminate the target, while the idler photons are non-locally modulated with a high-frequency cosine wave and never interact with the target. By leveraging the non-local modulation and the quantum correlation, the target detection is achieved with a precision of 0.64 $\pm$ 0.06 mm within one second over a measurement range of 2-8 m. As the acquisition time is up to 500 s, the system achieves a precision of 29 $\pm\ 4{\ \mathrm{μm}}$. Furthermore, our system realizes a 50 times precision improvement over the classical single-photon scheme in a background noise 37 dB stronger than the returned probe photons. With these advantages, our method will open venues for the development of high-precision, long-range, and noise-resilient target detection.

quant-ph↗