Searcharxiv⌕ Search

arXiv subjects

Pan Zhou

Publications and source records attributed to Pan Zhou.

At least 19 recordsLinked to original sources
arXiv2026

ScenePilot: Grow-and-Repair Policy for Text-Driven 3D Indoor Scene Generation

Text-driven 3D indoor scene generation has advanced from dataset-bound layout modeling to open-vocabulary synthesis with large language and vision-language models. Yet existing methods remain limited: one-pass generators often yield geometrically invalid layouts, heavy post-hoc optimization is costly and unstable, and prompt-only planners lack reusable layout priors for functional grouping and object relations. We propose \textbf{ScenePilot}, a retrieval-augmented \textbf{Grow-and-Repair} framework that formulates scene generation as prior-guided incremental growth with learned rectification. Given a prompt, the Hierarchical Retrieval-Augmented Planning (HRAP) module retrieves room-, group-, and anchor-level layout priors to support functional group planning. A text-driven base generator then inserts object groups sequentially, while the Reinforcement Multimodal Repair (RMR) module performs lightweight local correction after each insertion and a final global repair after completion. To train this policy, we construct \textbf{SceneReverse-17k}, a repair-trajectory dataset built by perturbing high-quality 3D scenes in position, rotation, and scale, then using inverse operations as executable rectification targets. The policy predicts structured \emph{move--rotate--scale} actions from rendered views, scene state, retrieved priors, and edit history. By combining HRAP with RMR, ScenePilot offers an efficient alternative to one-shot generation and heavy full-scene optimization, improving physical plausibility, functional coherence, and controllability while preserving diversity.

cs.CV↗
arXiv2026

Prediction of a layer nonlinear Hall effect in bilayer nonmagnetic or antiferromagnetic systems

Nonlinear Hall effects provide a powerful probe of quantum geometry in solids and enable rectification phenomena beyond the constraints of linear response. In this Letter, we predict a \emph{layer nonlinear Hall effect} (LNHE) in stacked bilayer systems composed of nonmagnetic or antiferromagnetic materials with a vanishing linear Hall conductivity. In such systems, the second- or third-order nonlinear Hall responses are intrinsically layer odd: the contributions from the two constituent layers have equal magnitude but opposite sign, resulting in exact cancellation under layer-exchange symmetry. An out-of-plane electric field $E_z$ can break this symmetry, thereby unveiling the hidden response and converting it into a switchable macroscopic nonlinear Hall signal. Using a minimal $k\!\cdot\!p$ model, we demonstrate that the LNHE can originate from the Berry curvature dipole mechanism. A systematic symmetry analysis of all 80 layer groups further yields a complete classification of the symmetry constraints and stacking configurations that allow for this type of LNHE. Beyond this mechanism, additional symmetry analysis reveals that LNHE may also arise from quantum metric dipole or inversed mass dipole. Remarkably, even in cases where second-order nonlinear Hall responses are symmetry forbidden, a third-order LNHE can still survive in certain stacked bilayer configurations. First-principles calculations on representative bilayers---nonmagnetic 1T$'$-WTe$_2$ and 1T$'$-ReS$_2$---explicitly demonstrate electrically reversible second-order LNHE, in full agreement with our symmetry-based predictions. Overall, our results establish LNHE as a universal phenomenon in a wide range of layered quantum materials and provide a robust route toward electrically tunable nonlinear transport.

cond-mat.mtrl-sci↗
arXiv2026

SubZero+: Efficient Zeroth-Order LLM Fine-Tuning via Large Learning Rates

Zeroth-order (ZO) optimization enables backpropagation-free fine-tuning of large language models, but existing ZO methods suffer from high-variance gradient estimators, making convergence unstable and highly sensitive to learning rates. We propose SubZero+, an improved SubZero framework that improves stability in three complementary ways: (i) multi-query gradient estimation within layer-specific low-rank subspaces to reduce variance without exhibiting the multi-query paradox; (ii) a subspace Adam optimizer that performs adaptive updates using in-subspace multi-query gradient statistics; and (iii) a sign correction for QR-based subspace construction to ensure Haar-distributed projection matrices, eliminating implementation-dependent orientation ambiguity. Experiments on models from 1.3B to 32B across SuperGLUE, under both full-parameter tuning and LoRA, show that SubZero+ consistently outperforms prior ZO baselines, enlarges the stable learning-rate range, and narrows the gap to first-order methods with minimal extra memory overhead.

cs.LG↗
arXiv2026

CardioState-JEPA: Delay-Aware Cross-Modal Learning of a Shared Cardiac Representation

Electrocardiography (ECG), photoplethysmography (PPG), and phonocardiography (PCG) provide complementary views of the same cardiac cycle, yet existing cardiac foundation models are trained for a single sensing modality, leaving the shared physiology across sensors unexploited. We introduce CardioState-JEPA, a cardiac foundation model to learn a single shared representation jointly across ECG, PPG, and PCG, built on a physiology-aware joint-embedding predictive architecture. The model maps heterogeneous waveforms into a common token space, processes them with a single shared Transformer encoder, and learns by predicting masked latent cardiac states, placing the pretraining target on shared physiology rather than sensor-specific waveform appearance. To handle the temporal offsets between electrical, mechanical, and hemodynamic events, cross-modal prediction uses a learned delay aligner that matches signals at the corresponding cardiac time. Because synchronized multi-sensor recordings are scarce, CardioState-JEPA first learns within-modality structure from abundant unimodal data and then uses paired data to align modalities in latent cardiac time. Evaluated as a frozen encoder across 25 downstream tasks spanning ECG, PPG, and PCG, our encoder improves average PPG classification by 8.2 AUROC points, PCG murmur detection by 18.8 AUROC points, and ECG classification by 15.5 AUROC points over the best self-supervised signal baseline and matches or exceeds cardiac models trained with privileged clinical text or supervised labels on several ECG benchmarks. These results establish that heterogeneous cardiac signals can mutually supervise a single foundation model of cardiac physiology.

cs.LG↗
arXiv2026

RISE-RL: Rubric-Informed Selective Exploration for Open-Ended Reinforcement Learning

Aligning Large Language Models (LLMs) for open-ended tasks is challenging because responses must satisfy multidimensional criteria without following a single correct generation trajectory. Existing rubric-based reinforcement learning (RL) methods compress fine-grained criterion-level feedback into scalar rewards, making persistent capability gaps difficult to target under limited on-policy exploration. We propose $\textbf{RISE-RL}$ (Rubric-Informed Selective Exploration), which uses repeatedly missed rubric criteria to elicit privileged trajectories that are difficult to discover through unguided exploration alone. RISE-RL retains only trajectories whose complete-rubric reward exceeds the mean reward of natural rollouts, and then re-evaluates them under the original prompt to emphasize behaviors that remain weakly supported by the natural policy. The resulting guidance signal is optimized through a separate auxiliary objective and removed once its additional benefit diminishes. Experiments with 4B and 14B models across writing, chat, health, and science show that RISE-RL achieves the highest mean score on every evaluated benchmark under guidance-free evaluation. Compared with standard Rubric-RL, it improves the average score by 1.3 points at the 4B scale and $\textbf{3.3 points at the 14B scale}$, including a $\textbf{6.0-point}$ gain on CreativeWriting-V3. It also improves creative-writing diversity and yields gains on objectively scored medical and scientific benchmarks. These results indicate that selective internalization through reward filtering and policy support shaping is effective for open-ended reinforcement learning.

cs.AI↗
arXiv2026

Symmetry Classification of Non-Relativistic Hidden Spin Polarization in Noncollinear Magnets

Hidden spin polarization (HSP), in which spin-polarized states exist locally while the total spin polarization are hidden in momentum space, has been extensively studied in nonmagnetic and collinear magnetic systems, but remains largely unexplored in noncollinear magnets. Here we establish a unified symmetry framework for HSP in noncollinear magnetic materials based on spin-group theory. We show that spin symmetries systematically constrain nonrelativistic spin polarization, giving rise to four distinct split spin-texture (SST) types for each local sector, denoted as SST-1, SST-2, SST-3, and SST-4. Based on these splitting forms, together with the dimensionality of the associated local spin textures and the symmetry relations between different local sectors, we further classify HSP into three categories: HSP-1, HSP-2, and HSP-3. We illustrate these categories using tight-binding models and representative material examples, including SrFe$_2$Se$_2$O, USb, Sr$_2$Mn$_3$Sb$_2$O$_2$, PrFeAsO, and GdMn$_2$Si$_2$. A survey of the MAGNDATA database further identifies 133, 7, and 139 candidate noncollinear magnetic materials hosting HSP-1, HSP-2, and HSP-3, respectively. In addition, our symmetry analysis and first-principles calculation show that many of these materials can exhibit nonzero spin-related response tensors. These results establish a general framework for understanding HSP in noncollinear magnets and highlight their potential for spin-dependent functionalities.

cond-mat.mtrl-sci↗
arXiv2026

Emergent Surface Altermagnetism

Research on altermagnetism has thus far primarily focused on spin-polarized bulk electronic states in magnetic materials. In this work, we advance the field by introducing the concept of surface altermagnetism (SAM), wherein altermagnetic spin polarization emerges at the surfaces of collinear antiferromagnets (AFMs) or altermagnets (AMs). To lay the theoretical groundwork for this phenomenon, we construct a thorough symmetry-based framework that systematically connects bulk spin groups to surface spin groups for both types of systems. Through symmetry analysis, we identify all symmetry-breaking surfaces capable of supporting SAM, identifying 35 for $PT$-symmetric AFMs and 61 distinct cases for bulk AMs. Moreover, we show that 203 collinear spin space groups---including 100 without and 103 with the $[C_2 \Vert P]$ operation---permit the appearance of SAM on the surface of $tT$-symmetric AFMs via the breaking of fractional translational symmetries. The proposed framework is verified using tight-binding models and first-principles calculations, with practical material implementations shown in representative compounds like NaMnP, LiMnAs, and CrSb. Our results establish SAM as a robust, symmetry-protected magnetic state, extending altermagnetic phenomena to material surfaces and paving the way for advanced, field-free spin manipulation in next-generation spintronic technologies.

cond-mat.mtrl-sci↗
arXiv2026

Hear, Invoke, and Understand: A Skill-Calling Multimodal Agent for Large Audio Language Models

Complex acoustic problems may require models to perform acoustic operations, interact with external tools and reason over the resulting textual or processed-audio observations rather than answer directly from a fixed audio input. We study such problems as tool-interactive audio reasoning and develop SpeechAgent-R, an audio agent that coordinates its intrinsic multimodal understanding with external skills and tools. To support this capability, we construct HIU-Corpus, comprising 65,492 interaction trajectories and 507.6 hours of audio across 24 tasks, 8 skills and 9 tools. SpeechAgent-R first learns structured interaction behaviors through trajectory-based supervised fine-tuning and then improves its decisions through multi-turn reinforcement learning. We further introduce HIU-Bench to jointly evaluate task performance, interaction quality and generalization to diverse task settings. It contains 1,395 samples across 56 tasks, including in-distribution (ID) and out-of-distribution (OOD) splits with substantial shifts in tool usage and workflow composition. SpeechAgent-R achieves 84.17 on ID tasks and 70.94 on OOD tasks, improving over the base model under the same agent harness by 15.40 and 14.23 points. These results demonstrate that learning skill and tool coordination improves audio agents' ability to handle diverse task settings and adaptive tool interactions.

cs.MM↗
arXiv2026

Layer Edelstein Effect

Electrical control of magnetism represents a fundamental route toward next-generation spintronic functionalities. In this Letter, we introduce a universal current-induced spin phenomenon in bilayer systems, termed the layer Edelstein effect (LEE), which serves as the natural counterpart of the layer Hall effect in real space. It is defined by the emergence of layer-resolved spin magnetizations with opposite components on the top and bottom layers, driven by an in-plane charge current and controllable by an external electric field. We establish the general existence of the LEE using a minimal bilayer $k \cdot p$ theory. By combining symmetry analysis with a general bilayer stacking framework, we derive a model-independent symmetry criterion demonstrating that the LEE is generically allowed in a broad class of nonmagnetic bilayer stacking systems. We further show that the LEE admits two universal manifestations: explicit layer-opposite spin magnetization components mandated directly by symmetry, and components become activated upon symmetry reduction by external electric fields. First-principles calculations on stacked bilayer MoSSe, MoTe$_2$ and WTe$_2$ confirm the predicted effect and illustrate their experimental feasibility. Our work establishes the LEE as a generic symmetry-governed response of bilayer systems, providing a unified conceptual framework for electrically generating and manipulating layer-resolved spin polarization.

cond-mat.mes-hall↗
arXiv2026

RFMSR: Residual Flow Matching for Image Super-Resolution

Image super-resolution (ISR) has witnessed remarkable progress with diffusion models and flow matching. The dominant text-to-image (T2I) based approaches leverage large-scale foundation models as generative priors, achieving impressive perceptual quality but at the cost of massive model sizes and prohibitive training expenses. Recent flow-matching-based vision-only approaches have made significant strides; however, they adopt standard flow formulations that transport from a pure Gaussian prior to the data distribution, discarding the rich structural information already present in the low-quality (LQ) input. Furthermore, existing single-step acceleration techniques often forfeit the model's multi-step inference capability. In this paper, we propose Residual Flow Matching for Image Super-Resolution (RFMSR), a vision-only framework that centers the source distribution at the LQ latent, reducing transport distance and preserving structural priors throughout the flow trajectory. We further introduce a two-phase training strategy: Phase I pretrains the velocity field via conditional flow matching, while Phase II applies end-to-end supervision to the single-step prediction while retaining the velocity loss across all timesteps, achieving high-quality single-step generation without sacrificing multi-step refinement. Extensive experiments demonstrate that RFMSR achieves comparable or even superior perceptual quality compared to state-of-the-art (SOTA) methods. The source code is available at https://github.com/Faze-Hsw/RFMSR.

cs.CV↗
arXiv2026

MSU-Bench: Towards Speaker-Centric Understanding in Conversational Multi-Speaker Scenarios

Spoken Language Understanding (SLU) is moving from task-specific pipelines toward large audio language models (LALMs) that generate natural-language responses. However, existing speech benchmarks mainly focus on single-speaker settings or isolated subtasks, leaving speaker-centric understanding in realistic multi-speaker conversations insufficiently evaluated. We introduce MSU-Bench, a diagnostic benchmark for multi-speaker conversational understanding, covering 16 speaker-centric tasks and 2,300 QA instances in a two-tier framework from speaker grounding to dialogue reasoning. We build a Gemini-assisted annotation and QA generation pipeline with human-in-the-loop verification, achieving high QA validity and strong agreement between human answers and verified labels. We further analyze speaker-referencing schemes and diagnostic error types to reveal bottlenecks in speaker grounding and reasoning. Experiments reveal clear gaps across model families, with closed-source systems leading overall but all models still facing challenges in complex speaker grounding and multi-speaker reasoning. The benchmark annotations, metadata, and evaluation scripts will be available at the GitHub repository: https://github.com/ASLP-lab/MSU-Bench.

eess.AS↗
arXiv2026

Latent Thought Flow: Efficient Latent Reasoning in Large Language Models

Large Language Models (LLMs) increasingly rely on intermediate reasoning, yet explicit Chain-of-Thought (CoT) suffers from a linguistic space bottleneck: each thought must be decoded into tokens, causing high inference overhead. Latent reasoning moves deliberation into continuous space, but existing methods mostly learn deterministic or reward-maximizing paths, lacking a principled way to allocate probability across trajectories with different correctness and costs. We propose Latent Thought Flow (LTF), which models reasoning as variable-length continuous trajectories and trains a sampler to match a reward-induced posterior over answer quality and computation cost. We instantiate this with a continuous GFlowNet using stochastic latent transitions. To handle sparse answer supervision, we introduce an Entropy-Weighted Subtrajectory Balance objective for intermediate rewards and a reference-prior regularizer to anchor exploration. Experiments under finetuning and transfer learning settings show that LTF outperforms explicit CoT and latent reasoning baselines, improving accuracy by 9.5% while reducing reasoning length by 27.2% on average compared with strong latent reasoning baselines.

cs.AI↗
arXiv2026

PathRelax: Parallel-Path Relaxed Speculative Jacobi Decoding for Accelerating Auto-Regressive Text-to-Image Generation

The growing need for high-resolution image generation in autoregressive text-to-image models has resulted in extended token sequences, significantly increasing computational costs and inference times. However, existing state-of-the-art methods for accelerating autoregressive text-to-image models rely on chain-structured draft token sequences, leading to inefficient draft token search and limited acceptance lengths. To address this, we propose parallel-path cross-relaxed speculative Jacobi decoding (\textbf{PathSpec}), a novel framework that enhances efficiency through a multi-sequence draft tree structure. Our parallel-path speculative Jacobi decoding (\textbf{PathExplore}) expands the token search space, achieving a higher speedup ratio without sacrificing image quality. Additionally, we introduce cross-path relaxed verification (\textbf{PathRelax}) that exploits semantic similarities across sequences to further boost token acceptance rates. Evaluated on the Parti-Prompts, MSCOCO2017, and T2ICompBench datasets, our method achieves a speedup ratio of 4.14 $\times$, 3.95$\times$, and 4.18$\times$, respectively. Remarkably, PathExplore, without any relaxed sampling, outperforms relaxed sampling methods in the speedup ratio, such as GSD and LANTERN. Moreover, PathRelax's relaxation mechanism can be seamlessly integrated with other relaxation techniques, enabling further acceleration and providing an efficient solution for real-time text-to-image generation. Our code is available at https://github.com/Haodong-Lei-Ray/PathSpec.

cs.CV↗
arXiv2026

LiAuto-GeoX: Efficient Grounded Driving Transformer

Dense 3D reconstruction has demonstrated immense potential for spatial understanding, yet its viability as a real-time, onboard representation for autonomous driving remains an open challenge. Existing large-scale visual geometry models typically require substantial computational resources and lack the long-range geometric fidelity, surround-view consistency, and real-time efficiency demanded by dynamic driving environments. To bridge this gap, we present \textbf{LiAuto-GeoX}, an efficient grounded driving transformer designed for deployable, ego-centric 3D scene understanding. Our approach begins by learning a high-capacity driving geometry model from large-scale surround-view data, utilizing sparse LiDAR priors to provide robust geometric grounding in distant, ambiguous, or structure-sparse regions. We then instantiate this capability into a highly compact 155M-parameter onboard model through a novel geometry-preserving distillation framework. This framework employs mask-guided depth-aware distillation to retain fine-grained metric structures by emphasizing geometrically informative regions, and relative-pose relational distillation to enforce cross-view spatial consistency through pose-induced geometric relations. Extensive evaluations reveal that \textbf{LiAuto-GeoX} runs at 220 FPS on KITTI while maintaining high-fidelity dense reconstruction, enabling real-time deployment. The learned geometry transfers seamlessly to downstream autonomy tasks, achieving 90.6 PDMS in trajectory prediction, 24.63 mIoU in occupancy prediction, and 47.67 IoU in future-frame prediction. These all demonstrate that efficient dense 3D reconstruction can transcend its traditional role as a perception target to serve as a scalable, foundational geometric representation for next-generation autonomous driving.

cs.CV↗
arXiv2026

D$^3$-MoE:Dual Disentangled Diffusion Mixture-of-Experts for Style-Controllable End-to-End Autonomous Driving

Traditional end-to-end autonomous driving frameworks frequently suffer from the "style-averaging" dilemma when trained on high-variance human demonstrations, yielding homogenized, style-uncontrollable, and even kinematically unsafe policies. To overcome this limitation, we present D$^3$-MoE (Dual Disentangled Diffusion Mixture-of-Experts), which disentangles trajectory modeling along two complementary axes. On the behavioral axis, generation is decoupled from selection: a style-conditioned diffusion process synthesizes multi-style candidate trajectories in parallel within a single scene, allowing a downstream module to select the optimal trajectory based on user preference or an evaluation score. On the physical axis, decoupled longitudinal and lateral routers activate their respective experts during inference time, trained without manual labels using self-supervised targets from orthogonal ground-truth kinematics. These activated experts, architected as Diffusion Transformers (DiT) and equipped with style-conditioned AdaLN and asymmetric lateral-fusion cross-attention, independently predict their corresponding physical state before being reassembled into a unified, kinematically coherent trajectory. Extensive evaluations on the challenging NAVSIM benchmark demonstrate that D$^3$-MoE achieves state-of-the-art planning performance, reaching 88.2 PDMS and 84.3 EPDMS by default. Moreover, our Best-of-Three ensemble strategy effectively broadens the multi-modal solution space, raising performance to 91.3 PDMS and 87.5 EPDMS. Both quantitative and qualitative analyses jointly confirm the framework's advantages in planning quality and style controllability.

cs.RO↗
arXiv2026

HMPO: Hybrid Median-length Policy Optimization for Chain-of-Thought Compression

Large language models achieve remarkable performance via extended chain-of-thought (CoT) reasoning, yet this lengthy process incurs substantial inference overhead. Existing CoT compression methods struggle with inflexible manual length budgets, computationally expensive multi-stage training pipelines, and fragile scalability restricted to small models. We propose HMPO (Hybrid Median-length Policy Optimization), a cost-effective, single-stage reinforcement learning framework. HMPO efficiently compresses CoT via three synergistic components: an adaptive median-based budget derived from successful rollouts to eliminate manual tuning, a cosine-decay token reward for smooth length penalization, and a multiplicative reward formulation that substantially mitigates trivial reward hacking by strictly prioritizing answer correctness. Trained exclusively on mathematical data, HMPO generalizes seamlessly across math, code, science, and instruction-following tasks. Extensive experiments scaling from 9B to 122B parameters across dense and Mixture-of-Experts (MoE) architectures demonstrate that HMPO achieves 19%--46% token compression with negligible accuracy degradation, all while drastically reducing training costs compared to existing multi-stage baselines.

cs.LG↗
arXiv2026

Stacking-Engineered Switchable Altermagnetism in Topological FeSe bilayer systems

Altermagnetism and topological insulators represent two of the most transformative frontiers in modern condensed matter physics, spintronics, and quantum information science. Bringing these two paradigms together opens a largely unexplored route toward fundamentally new quantum phenomena. Here, we predict a topological altermagnetic phase in bilayer tetragonal Fe-based superconductors and reveal it as a highly tunable platform for valley-polarized anomalous Hall physics. Based on first-principles calculations, we show that the characteristic spin-splitting and valley polarization can be effectively tuned via applied strain. Moreover, the resulting valley-polarized anomalous Hall conductivity can be manipulated by shifting the Fermi level. These findings reveal a powerful route for controlling altermagnetism in topological materials and identify a realistic material platform for its experimental realization and technological exploitation.

cond-mat.mtrl-sci↗
arXiv2026

Annotations Are Not All You Need: A Cross-modal Knowledge Transfer Network for Unsupervised Temporal Sentence Grounding

This paper addresses the task of temporal sentence grounding (TSG). Although many respectable works have made decent achievements in this important topic, they severely rely on massive expensive video-query paired annotations, which require a tremendous amount of human effort to collect in real-world applications. To this end, in this paper, we target a more practical but challenging TSG setting: unsupervised temporal sentence grounding, where both paired video-query and segment boundary annotations are unavailable during the network training. Considering that some other cross-modal tasks provide many easily available yet cheap labels, we tend to collect and transfer their simple cross-modal alignment knowledge into our complex scenarios: 1) We first explore the entity-aware object-guided appearance knowledge from the paired Image-Noun task, and adapt them into each independent video frame; 2) Then, we extract the event-aware action representation from the paired Video-Verb task, and further refine the action representation into more practical but complicated real-world cases by a newly proposed copy-paste approach; 3) By modulating and transferring both appearance and action knowledge into our challenging unsupervised task, our model can directly utilize this general knowledge to correlate videos and queries, and accurately retrieve the relevant segment without training. Extensive experiments on two challenging datasets (ActivityNet Captions and Charades-STA) show our effectiveness, outperforming existing unsupervised methods and even competitively beating supervised works.

cs.CV↗