arXiv subjects

Wei Zhu

Publications and source records attributed to Wei Zhu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Layer-engineered quantum anomalous Hall effect in twisted rhombohedral graphene

Realizing programmable topological states in quantum anomalous Hall (QAH) insulators requires the ability to design and dynamically tune their topological invariant, the Chern number C. Here, we report a designer QAH platform based on twisted rhombohedral graphene family, in which C becomes a programmable and electrically tunable degree of freedom. By engineering the layer configuration in twisted monolayer-rhombohedral N-layer graphene, denoted as (1+N)L, we realize QAH states with C=N at moire filling v=1, where the layer number N=3,4,5 directly sets the Chern number. Beyond such static layer programming, we demonstrate in-situ electrical control. In a twisted monolayer-trilayer device, the sign of C (chirality) can be switched by electrostatic doping or displacement field. Most strikingly, in twisted Bernal bilayer-rhombohedral tetralayer graphene denoted as (2+4)L, we drive a displacement-field-induced topological phase transition between two distinct QAH states with C=3 and C=4 in a single device. Our work establishes a layer-engineered and electrically tunable platform that transitions topological quantum matter from discovery to design, opening the way toward on-demand engineering of correlated topological states and reconfigurable topological electronics.

cond-mat.mes-hall↗

arXiv2026

Conformal Nature of Quantum Phase Transitions via Fuzzy Three-Sphere Regularization

Conformal field theory (CFT) offers a modern viewpoint for understanding phase transitions. However, directly accessing the conformal algebra and microscopically uncovering the emergent conformal symmetry, especially in higher dimensions, remains a significant challenge. Motivated by recent advances in revealing CFT features via the fuzzy two-sphere, here we generalize this approach to higher dimensions and aim to expose the conformality at the (3+1)-D quantum critical point. We demonstrate this framework by investigating quantum phase transitions belonging to the Ising and Yang-Lee universality classes in a (3+1)-D quantum model (equivalent to a classical four-dimensional system), realized via Landau level projection on the fuzzy three-sphere. By computing the energy spectra at criticality, we explicitly verify the state-operator correspondence, a hallmark of conformal invariance. Together with prior advances, this work establishes a new pathway for the microscopic study of emergent conformality in higher-dimensional phase transitions.

cond-mat.str-el↗

arXiv2026

Inverse Low-Dimensional Manifold Reconstruction Framework for Spatiotemporal Reconstruction of Compressible Physical Fields

Compressible physical fields are widely present in the real physical world, but current artificial intelligence lacks an understanding mechanism for the non-differentiable features in compressible physical fields. Addressing the limitations of existing deep learning architectures in handling global non-differentiable features, we propose the Inverse Low-Dimensional Manifold reconstruction framework (ILDM). This framework couples the Non-differentiable Approximation Function (NAF) for capturing non-differentiable features in compressible flows with the Smooth Fluid Reconstruction (SFR) module tailored for smooth fluid regions. Extensive evaluations across 1D and 2D benchmarks, including Riemann problems and double Mach reflection, demonstrate that ILDM significantly outperforms cPINN and R-adaptive DeepONet. Specifically, ILDM achieves superior localization of non-differentiable interfaces and maintains robust super-resolution performance even with low-resolution inputs, establishing a physically consistent and scalable paradigm for data-driven fluid dynamics.

physics.comp-ph↗

arXiv2026

Learning Lax Pairs: Revisiting the Classical Paradigm

A Lax pair $(L,P)$ is sometimes thought of as a structural certificate, in that the spatial operator $L$ carries the spectral data of an integrable system, and its isospectral evolution under $\partial_t L = [L,P]$ encodes the nonlinear dynamics. Yet, experience shows that the correspondence between equations and Lax pairs is much more nuanced than this picture suggests. Equations can admit Lax pairs that fail to encode the expected integrable structure. This paper probes that anomalous corner of the Lax pair landscape through five case studies (the Euler top, the free Schrödinger equation, the inviscid Burgers equation, the shallow water system, and the Korteweg--de Vries equation), each illustrating a different way the link to integrability can be distorted. The approach combines analytical calculations with the Sparse Identification of Lax Operators (SILO) framework, which proved useful throughout, in some cases confirming the textbook pair and in others surfacing alternatives worth understanding on their own terms. The recurring lesson across the five cases is that compatibility underdetermines the Lax representation, so that anomalous pairs are regular features of the landscape rather than pathologies. Notably, we show that a spectrally degenerate Korteweg--de Vries Lax pair, classified as fake by standard criteria, still generates the full conservation hierarchy through its operator algebra, which shows that a blunt dichotomy between true and fake Lax pairs can be too reductive.

nlin.SI↗

arXiv2026

TESS planets in known radial velocity cold Jupiter systems: Hot super Earth occurrence is enhanced by cold Jupiters

The correlation between inner super-Earths (SEs) and outer cold Jupiters (CJs) provides an important constraint on the formation and dynamical evolution of planetary systems. Previous studies have suggested a positive connection between these two populations, particularly around metal-rich stars, and proposed that nearly all CJ-hosting stars may also harbor inner SEs. In this work, we use TESS transits to investigate the occurrence of hot super-Earths (HSE; $1$-$4R_\oplus$, $P<10\mathrm{d}$) in systems with known CJs detected by radial velocity. Out of a statistical sample of 132 CJ systems, we identify five transiting HSEs around four stars, including one new candidate (TOI-6965.01) around HD 50554. To enable statistical analysis, we first validate the two candidates around HD 50554 using TESS photometry, archival RV measurements, and Gaia astrometry. After accounting for detection sensitivity and geometric transit probability, we find that the presence of CJs enhances the occurrence rate of HSEs by a factor of $6.5^{+3.1}_{-2.3}$ relative to field stars, with the case of no enhancement being ruled out at 99.9% confidence level. Taking into account the average multiplicity of HSEs, we find that about 91% of CJ systems host at least one HSE. Our results provide strong supporting evidence for a positive HSE-CJ correlation. We also briefly explore the correlation around metal-poor hosts and for specific sub-populations (e.g., warm super Earths or cold super Jupiters).

astro-ph.EP↗

arXiv2026

Generalization Guarantees for Multi-Input Neural Operator Learning in Sobolev Spaces

We develop approximation and generalization error estimates for multi-input neural operators, with the output error measured in Sobolev norms. In contrast to standard operator-learning settings with a single input function, our framework allows multiple input functions defined on possibly different domains, with different dimensions and Sobolev regularities. The derived rates explicitly quantify the contribution of each input space to the final error bound. In particular, in the balanced regime, the approximation and generalization rates are governed by the interaction between the input dimensions, regularities, and Sobolev orders, while the dependence on the model complexity retains a $\log\log/\log$-type structure. Our analysis provides a general theoretical framework for multi-input operator learning, including Sobolev training, and is applicable to operator learning problems arising from partial differential equations and scientific computing.

cs.LG↗

arXiv2026

FireRed-Image-Edit-1.0 Technical Report

We present FireRed-Image-Edit, a diffusion transformer for instruction-based image editing that achieves state-of-the-art performance through systematic optimization of data curation, training methodology, and evaluation design. We construct a 1.6B-sample training corpus, comprising 900M text-to-image and 700M image editing pairs from diverse sources. After rigorous cleaning, stratification, auto-labeling, and two-stage filtering, we retain over 100M high-quality samples balanced between generation and editing, ensuring strong semantic coverage and instruction alignment. Our multi-stage training pipeline progressively builds editing capability via pre-training, supervised fine-tuning, and reinforcement learning. To improve data efficiency, we introduce a Multi-Condition Aware Bucket Sampler for variable-resolution batching and Stochastic Instruction Alignment with dynamic prompt re-indexing. To stabilize optimization and enhance controllability, we propose Asymmetric Gradient Optimization for DPO, DiffusionNFT with layout-aware OCR rewards for text editing, and a differentiable Consistency Loss for identity preservation. We further establish REDEdit-Bench, a comprehensive benchmark spanning 15 editing categories, including newly introduced beautification and low-level enhancement tasks. Extensive experiments on REDEdit-Bench and public benchmarks (ImgEdit and GEdit) demonstrate competitive or superior performance against both open-source and proprietary systems. To support future research, our code, models, and benchmark suite are publicly available at https://github.com/FireRedTeam/FireRed-Image-Edit/ .

cs.CV↗

arXiv2026

ReSkill: Reconciling Skill Creation with Policy Optimization in Agentic RL

Agentic reinforcement learning (RL) enables LLM agents to improve continuously from environment rewards, yet the resulting policies do not systematically accumulate reusable strategies that generalize across tasks. Modular skills can provide such reusable strategies, yet existing skill-augmented RL methods decouple skill creation from policy optimization, risking adopting skills that conflict with the evolving policy. Inspired by Anthropic's Skill Creator, we introduce ReSkill, an RL-in-the-loop skill creation framework that reconciles skill evolution with policy learning. ReSkill exploits the group-wise structure of GRPO to naturally embed three mechanisms with only marginal additional overhead: (1) an assertion-driven skill creator that diagnoses failures from past experience and proposes conditional, trigger-based skill revisions; (2) within-group rollout sampling that enables controlled comparison of skill versions, capturing which version best supports the policy's ongoing learning; and (3) Thompson Sampling with adaptive discounting to balance exploration and exploitation in skill version selection as the policy evolves. Across several domains, ReSkill consistently outperforms existing memory and skill-based RL methods, with the largest gains on unseen tasks. Analysis of the skill lifecycle shows skills being automatically created, tested, refined, and pruned as the policy improves, demonstrating reconciled skill-policy co-evolution.

cs.AI↗

arXiv2026

TRANS: Terrain-aware Reinforcement Learning for Agile Navigation of Quadruped Robots under Social Interactions

This study introduces TRANS: Terrain-aware Reinforcement learning for Agile Navigation under Social interactions, a deep reinforcement learning (DRL) framework for quadrupedal social navigation over unstructured terrains. Conventional quadrupedal navigation typically separates motion planning from locomotion control, neglecting whole-body constraints and terrain awareness. On the other hand, end-to-end methods are more integrated but require high-frequency sensing, which is often noisy and computationally costly. In addition, most existing approaches assume static environments, limiting their use in human-populated settings. To address these limitations, we propose a two-stage training framework with three DRL pipelines. (1) TRANS-Loco employs an asymmetric actor-critic (AC) model for quadrupedal locomotion, enabling traversal of uneven terrains without explicit terrain or contact observations. (2) TRANS-Nav applies a symmetric AC framework for social navigation, directly mapping transformed LiDAR data to ego-agent actions under differential-drive kinematics. (3) A unified pipeline, TRANS, integrates TRANS-Loco and TRANS-Nav, supporting terrain-aware quadrupedal navigation in uneven and socially interactive environments. Comprehensive benchmarks against locomotion and social navigation baselines demonstrate the effectiveness of TRANS. Hardware experiments further confirm its potential for sim-to-real transfer.

cs.RO↗

arXiv2026

Looking for Condensed Gluons: A Cross-Scale Journey from the Deep Structure of Protons to High-Energy Cosmic Rays -- A Mini-Review

Quark-gluon dynamics within protons and high-energy radiation phenomena in the universe are typically regarded as two entirely distinct fields. This paper aims to demonstrate that gluon condensation (GC) may serve as a direct bridge between these two fields. We review three key aspects of GC research: first, the Zhu-Shen-Ruan (ZSR) equation, as a nonlinear evolution equation based on structural symmetry, exhibits self-consistent connections with the DGLAP, BFKL and GLR-MQ-ZRS equations, providing a theoretical foundation for the generation of GC; second, the chaotic solutions and the shadowing-antishadowing synergy inherent in this equation can drive gluons to aggregate near the critical momentum, thereby forming a novel type of high-density, strongly interacting matter; third, these changes in microstructure manifest themselves as a broken-power-law feature in high-energy cosmic gamma-ray spectra, thereby offering new insights into the hadronic scenarios underlying certain astrophysical sources. Consequently, GC not only concerns the novel behaviour of quantum chromodynamics under extreme conditions but may also serve as a vital window for probing the deep structure of protons using cosmic-ray signals. With the advancement of higher-precision gamma-ray observations, hadron collision experiments and related theoretical research, the physical picture of GC and its observational criteria are expected to undergo more rigorous testing. Should this picture be confirmed, certain features in the high-energy gamma-ray spectrum will need to be re-examined within the deeper context of hadronic dynamics; simultaneously, GC may also provide a new entry point for research into pion condensation in nuclear physics and even condensed matter physics. Consequently, the significance of the search for GC extends beyond the model itself, reaching into multiple fields of natural science.

hep-ph↗

arXiv2026

EditCaption: Human-Refined SFT and HAE-DPO for Image Editing Instruction Synthesis

High-quality source-target image pairs with precise editing instructions are essential for instruction-guided image editing, yet constructing such training triplets at scale remains costly. Recent pipelines often rely on vision-language models to synthesize editing instructions automatically, but we find that strong VLMs still struggle to describe visual transformations between image pairs. In particular, they exhibit three recurring failure modes: orientation inconsistency, viewpoint ambiguity, and missing fine-grained attributes. In a human evaluation on 400 image pairs, several open-source VLM baselines produce critical-error rates above 47\%, making many synthesized instructions unsuitable for downstream training. To address this, we propose EditCaption, a two-stage post-training pipeline for image editing instruction synthesis. First, we construct a 100K supervised fine-tuning dataset through GLM-based auto-captioning, EditScore filtering, and human refinement. Second, we collect 10K human-annotated preference pairs, where each rejected instruction is labeled with its primary error type and severity. Based on this dataset, we propose Hardness-Adaptive Error-Aware DPO (HAE-DPO), a task-adapted DPO objective that introduces an adaptive margin based on human-labeled severity, failure-mode type, and reference-model hardness. Experiments across three benchmarks demonstrate that our 235B model with SFT+HAE-DPO achieves state-of-the-art performance among open-source and closed models, scoring 4.720 on Eval-400, 4.672 on HQ-Edit, and 4.651 on ByteMorph-Bench -- surpassing Gemini-3-Pro on all three. Human evaluation confirms critical error rates drop from 47.75\% to 17.50\%, with correct rates improving from 41.75\% to 70.25\%, surpassing Gemini-3-Pro (66.00\%).

cs.CV↗

arXiv2026

Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training

Post-training has become essential for adapting large language models (LLMs) to complex downstream behaviors, including instruction following, preference alignment, and multi-step reasoning. Reinforcement learning with verifiable rewards (RLVR) has recently emerged as a particularly effective post-training paradigm for improving reasoning capabilities, with critic-free algorithms such as GRPO and GSPO enabling scalable optimization. However, RLVR post-training with full fine-tuning (FFT) requires substantial GPU memory and incurs high training costs. Although parameter-efficient fine-tuning (PEFT) methods, such as Low-Rank Adaptation (LoRA), effectively reduce computational costs, they often suffer from a noticeable performance gap compared to full fine-tuning in post-training for complex reasoning tasks. In this paper, we propose Hybrid-LoRA, an efficient hybrid post-training framework that selectively applies full fine-tuning to a small subset of modules less suited to low-rank adaptation, while adapting the remaining components with LoRA. We introduce a novel Hybrid-LoRA Score to rank candidate modules according to their sensitivity to low-rank adaptation under a fixed parameter budget. Experiments show that Hybrid-LoRA closely matches full fine-tuning performance under a 10% full fine-tuning module budget, with the remaining candidate modules adapted by LoRA, consistently outperforming four state-of-the-art PEFT post-training baselines, achieving improvements of up to 5.65% and on average 4.36% over the best baseline.

cs.LG↗

arXiv2026

Cascade of fractional quantum Hall states in 2D system

The observation of the fractional quantum Hall (FQH) effect in 2D electron gases ushered in investigations of topological phases driven by strong electron correlations. Their remarkable features include fractionalized elementary excitations, gapless boundary states, and non-trivial quantum entanglement patterns. Thanks to persistent efforts in the building of new platforms and making higher-quality samples, a diverse plethora of FQH states have been unveiled in experiments. We report a systematic study of ultrahigh-quality GaAs/AlGaAs quantum wells with mobility up to 3.7*10^7 cm^2/V/s using quantum transport measurements in nuclear adiabatic demagnetization and dilution refrigerators down to 1 mK. In addition to many FQH states that have already been identified in previous work, new longitudinal resistance dips are observed at filling factors 17/33 and 15/31. The application of an in-plane magnetic field causes disparate variations of the FQH states. The theoretical foundation of these states is discussed in the framework of composite fermion theory. While most fractions can be explained as non-interacting composite fermions forming integer quantum Hall states, a few states correspond to FQH states of composite fermions that arise from residual interaction between them. We summarize the observed fractions in the range of 0 < ν < 2 and propose a pattern to account for their experimental appearance that provides an intuitive picture about the relative strengths of different FQH states.

cond-mat.mes-hall↗

arXiv2026

Noise-Started One-Step Real-World Super-Resolution via LR-Conditioned SplitMeanFlow and GAN Refinement

Pre-trained text-to-image (T2I) diffusion models have shown strong potential for real-world image super-resolution (Real-ISR), owing to their noise-started generation process that enables realistic texture synthesis and captures the one-to-many nature of super-resolution. However, diffusion-based Real-ISR methods still face a fundamental efficiency-quality trade-off. Multi-step methods generate high-quality results by iteratively denoising random Gaussian noise under LR conditioning, but suffer from slow sampling. Recent one-step methods greatly improve efficiency, yet they typically replace noise-started generation with direct LR-to-HR restoration, which weakens stochasticity and limits realistic detail synthesis. To address this issue, we propose SMFSR, a noise-started one-step Real-ISR framework via LR-conditioned SplitMeanFlow and GAN refinement. SMFSR preserves the random-noise starting point of diffusion models and learns a direct noise-to-HR mapping conditioned on the LR image. To this end, Interval Splitting Consistency distills the multi-step generative trajectory into a single average-velocity prediction, enabling efficient one-step generation. To compensate for the reduced opportunity for progressive refinement, we further introduce a GAN refinement stage, where a DINOv3-based discriminator enhances realistic texture synthesis and variational score distillation aligns the generated outputs with the natural image distribution under a frozen diffusion teacher. Extensive experiments demonstrate that SMFSR achieves state-of-the-art perceptual quality among one-step diffusion-based Real-ISR methods while retaining fast single-step inference.

cs.CV↗

arXiv2026

IdGlow: Dynamic Identity Modulation for Multi-Subject Generation

Multi-subject image generation requires seamlessly harmonizing multiple reference identities within a coherent scene. However, existing methods relying on rigid spatial masks or localized attention often struggle with the "stability-plasticity dilemma," particularly failing in tasks that require complex structural deformations, such as identity-preserving age transformation. To address this, we present IdGlow, a mask-free, progressive two-stage framework built upon Flow Matching diffusion models. In the supervised fine-tuning (SFT) stage, we introduce task-adaptive timestep scheduling aligned with diffusion generative dynamics: a linear decay schedule that progressively relaxes constraints for natural group composition, and a temporal gating mechanism that concentrates identity injection within a critical semantic window, successfully preserving adult facial semantics without overriding child-like anatomical structures. To resolve attribute leakage and semantic ambiguity without explicit layout inputs, we further integrate a badcase-driven Vision-Language Model (VLM) for precise, context-aware prompt synthesis. In the second stage, we design a Fine-Grained Group-Level Direct Preference Optimization (DPO) with a weighted margin formulation to simultaneously eliminate multi-subject artifacts, elevate texture harmony, and recalibrate identity fidelity towards real-world distributions. Extensive experiments on two challenging benchmarks -- direct multi-person fusion and age-transformed group generation -- demonstrate that IdGlow fundamentally mitigates the stability-plasticity conflict, achieving a superior Pareto balance between state-of-the-art facial fidelity and commercial-grade aesthetic quality.

cs.CV↗

arXiv2026

Edit Where You Mean: Region-Aware Adapter Injection for Mask-Free Local Image Editing

Large diffusion transformers (DiTs) follow global editing instructions well but consistently leak local edits into unrelated regions, because joint-attention architectures offer no explicit channel telling the network where to apply the edit. We introduce AdaptEdit, a co-trained, instruction- and region-aware adapter framework that retro-fits a frozen DiT into a precise local editor without modifying its backbone weights. A lightweight Block Adapter at every transformer block injects a structured condition stream that factorizes what to edit (instruction semantics) from where to edit (spatial mask); a learned SpatialGate routes the adapter signal selectively into the edit region while keeping the rest of the image near-identical to the source; and a Region-Aware Loss focuses the training objective on the changing pixels. Because these components make the backbone's internal representation mask-aware end-to-end, a thin MaskPredictor head trained jointly with the editor can ground the edit region directly from the instruction and source image -- eliminating any user-mask requirement at deployment. We evaluate on two complementary benchmarks: MagicBrush (paired ground-truth targets) to measure pixel-level preservation and edit accuracy, and Emu-Edit Test (no ground-truth images, 9 diverse edit categories) to stress-test instruction following and generalization across edit types. On both, AdaptEdit achieves state-of-the-art results, simultaneously outperforming mask-free and oracle-mask baselines. A seven-variant ablation cleanly isolates the contribution of each component.

cs.CV↗

arXiv2026

The Intrinsic Multiplicity Distribution of Exoplanets Revealed from the Radial Velocity Method. II. Constraints on Giant Planet Multiplicity from Different Surveys

Compared to the commonly used planet occurrence rates, the multiplicity distribution of planets can be more useful in constraining the formation and evolution pathways of planetary systems. This work follows an earlier work of Zhu (2022) and derive the intrinsic multiplicity distribution of giant planets (with masses above Saturn mass) from two independent radial velocity (RV) surveys. In particular, we find that $(7.8\pm1.4\%, 2.3\pm1.2\%, 0.5^{+0.8}_{-0.3}\%)$ of Sun-like stars in the HARPS sample have $(1, 2, 3)$ giant planets within 10\,au, whereas $(7.3\pm2.8\%, 7.2\pm2.3\%, <1.3\%, 1.0^{+1.0}_{-0.6}\%)$ of Sun-like stars in the California Legacy Survey (CLS) have $(1, 2, 3, 4)$ giant planets within 10\,au. Here we have further cleaned the CLS sample and removed planet detections that were not discovered in the survey mode. The total fraction of Sun-like stars with at least one giant planet within 10\,au from the two samples are $10.6\pm1.2\%$ and $15.8\pm2.1\%$, respectively, and the difference may be accounted for by their different metallicity distributions. We briefly discuss the theoretical implications of our results. In particular, the inferred giant planet multiplicity distribution is inconsistent with most of the proposed theoretical models involving planet--planet scatterings, which predict too many multi-giant systems.

astro-ph.EP↗

arXiv2026

Dissecting Failure Dynamics in Large Language Model Reasoning

Large Language Models (LLMs) achieve strong performance through extended inference-time deliberation, yet how their reasoning failures arise remains poorly understood. By analyzing model-generated reasoning trajectories, we find that errors are not uniformly distributed but often originate from a small number of early transition points, after which reasoning remains locally coherent but globally incorrect. These transitions coincide with localized spikes in token-level entropy, and alternative continuations from the same intermediate state can still lead to correct solutions. Based on these observations, we introduce GUARD, a targeted inference-time framework that probes and redirects critical transitions using uncertainty signals. Empirical evaluations across multiple benchmarks confirm that interventions guided by these failure dynamics lead to more reliable reasoning outcomes. Our findings highlight the importance of understanding when and how reasoning first deviates, complementing existing approaches that focus on scaling inference-time computation.

cs.AI↗