arXiv subjects

Lei Yang

Publications and source records attributed to Lei Yang.

At least 37 records · Page 2Linked to original sources

arXiv2026

HelloWorld: Towards Practical Applications of Generative Driving World Models

Driving world models provide a promising route toward scalable counterfactual data generation and interactive simulation beyond recorded driving logs. Realizing this potential requires a system that can generalize across diverse scenes, respond faithfully to prescribed controls, generate coherent multi-sensor observations, and operate efficiently under repeated inference. We present \textbf{HelloWorld}, a 2B driving world model system designed around these requirements. HelloWorld progressively specializes broad visual and motion priors from heterogeneous video data into controllable driving generation using ego pose, HD maps, and 3D boxes. A block-causal generation interface, together with adaptation to self-generated context, aligns the model with sequential simulation. The system further supports synchronized seven-camera RGB generation and conditional LiDAR synthesis, and is distilled toward few-step inference for efficient deployment. Experiments evaluate visual quality, control fidelity, cross-view consistency, robustness under repeated generation, inference efficiency, and LiDAR synthesis. Together, HelloWorld provides a unified framework for scalable driving data generation and interactive simulation.

cs.CV↗

arXiv2026

A Very Big Video Reasoning Suite

Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy and over one million video clips, approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first large-scale scaling studies of video reasoning and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark toolkit, and models are publicly available at https://video-reason.com/?v=vbvr .

cs.CV↗

arXiv2026

Probing the $^{12}$C+$^{12}$C fusion reaction via zero-degree spectator measurement in the $^{12}$C($^{14}$N,$αd$)$^{20}$Ne quasi-free reaction

The 12C+12C fusion reaction is a key physical process in stellar evolution and supernova explosions. It not only determines the late evolutionary fate of massive stars but also directly influences the critical conditions for triggering Type Ia supernovae in accreting white dwarfs. In this work, the THM was employed to investigate the 12C(12C,a0)20Ne reaction channel of the 12C+12C fusion process, using 14N as the Trojan horse nucleus. Telescope detectors were placed at 0 and 15 deg. to design two experimental configurations covering the forward-angle regions where spectator particles are most likely to emerge. By applying the DWBA, two sets of astrophysical S*(E) factors for the two-body reaction 12C(12C,a0)20Ne were extracted from the three-body reaction 12C(14N,da0)20Ne and normalized to existing experimental data. The results show that, limited by the overall experimental resolution, the present study cannot resolve fine resonance structures. Within the astrophysical energy region of 0.5-2 MeV, the extracted S*(E) factor exhibits an increasing trend toward lower energies. The S*(E) factor obtained with the 0-deg configuration shows a flatter trend than that obtained with the 15-deg configuration. Supported by the quasi-free reaction simulation results, the divergence between the two data sets may reflect a combination of experimental acceptance effects, finite detector resolution, and possible differences in the relative contributions of reaction mechanisms. This study provides a systematic examination of the experimental design, quasi-free event selection strategy, and interpretation of the underlying physical mechanisms, serving as a useful reference for understanding the role of the 12C+12C fusion reaction in astrophysical processes.

nucl-ex↗

arXiv2026

Radiance-Field Guided Pretraining: Scaling Localization Models with Unlabeled Wireless Signals

Radio frequency (RF)-based indoor localization offers significant promise for applications such as indoor navigation, augmented reality, and pervasive computing. While deep learning has greatly enhanced localization accuracy and robustness, existing localization models still face major challenges in cross-scene generalization due to their reliance on scene-specific labeled data. To address this, we introduce Radiance-Field Reinforced Pretraining (RFRP). This novel self-supervised pretraining framework couples a large localization model (LM) with a neural radio-frequency radiance field (RF-NeRF) in an asymmetrical autoencoder architecture. In this design, the LM encodes received RF spectra into latent, position-relevant representations, while the RF-NeRF decodes them to reconstruct the original spectra. This alignment between input and output enables effective representation learning using large-scale, unlabeled RF data, which can be collected continuously with minimal effort. To this end, we collected RF samples at 7,327,321 positions across 100 diverse scenes using four common wireless technologies--RFID, BLE, WiFi, and IIoT. Data from 75 scenes were used for training, and the remaining 25 for evaluation. Experimental results show that the RFRP-pretrained LM reduces localization error by over 40% compared to non-pretrained models and by 21% compared to those pretrained using supervised learning.

cs.IT↗

arXiv2026

onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction

We present onPanda, an interactive tool for efficiently annotating LLM alignment data and agent trajectories. onPanda adopts token-level correction as its core interaction: while reading a model response, the annotator locates the first inappropriate token and either picks a substitute from the model's candidate tokens or types the correct text via free-form editing. The system then truncates everything after that position and continues generation from the corrected prefix, repeating this locate-correct-continue loop until a satisfactory response is obtained. This mechanism lets annotators precisely steer model outputs at low cost: a small controlled study suggests that onPanda reduces median annotation time by 52% over manual post-editing. Since the vast majority of tokens in the final response are generated by the model itself, the resulting data largely preserves the model's sampling distribution and is well suited for constructing on-policy SFT and preference data. Furthermore, the token-level corrections recorded during annotation provide fine-grained supervision with precise positions and naturally paired positive--negative samples. onPanda also connects to external tools and harnesses, enabling interactive trajectory annotation in realistic environments. In addition, we release Panda-CVL, a dataset annotated with onPanda, together with a benchmark for token-level correction.

cs.CL↗

arXiv2026

Towards robust multimodal 3D object detection via visual foundation models

Multimodal 3D object detection is fundamental to robust perception in autonomous driving because it integrates complementary information from LiDAR and camera sensors. However, existing methods often fail to maintain robustness under out-of-distribution (OOD) corruptions caused by sensor noise, adverse weather, and environmental changes. To address this problem, we propose RoboDistill, a robust and generalizable multimodal 3D object detection framework that leverages visual foundation models (VFMs), such as the Segment Anything Model (SAM). First, we introduce SAM-AD, a domain-specific pretraining strategy that fine-tunes SAM on autonomous-driving imagery to extract feature representations with rich semantic information. Second, we design the AD Feature Pyramid Network (AD-FPN) to refine and upsample SAM features at multiple scales for seamless fusion with LiDAR features. Third, we develop the Depth-Guided Wavelet Attention (DGWA) module, which suppresses high-frequency sensor noise while preserving critical contextual information. Finally, we introduce KD Fusion, in which the pretrained SAM-AD serves as a teacher that distills high-quality visual knowledge into a lightweight point-cloud network, thereby improving robustness under noisy conditions. Extensive experiments across 27 challenging OOD corruption settings show that RoboDistill generally delivers stronger or competitive detection performance and robustness relative to representative state-of-the-art methods. This work bridges the gap between VFMs and 3D object detection and advances robust multimodal perception for real-world autonomous-driving applications.

cs.CV↗

arXiv2026

VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning

Native visual reasoning treats visual generation as the medium of reasoning itself: visual states (i.e. images and videos) are not merely inputs to be understood or outputs to be rendered, but first-class substrates for problem solving beyond language. Yet progress remains bottlenecked by the lack of scalable training tasks, reliable feedback, and controlled comparisons across generative substrates. In this work, we introduce VBVR-Pro, a closed-loop testbed that makes native visual reasoning through generation trainable, verifiable, optimizable, and experimentally controllable. 1) Task scaling. VBVR-Pro turns visual reasoning into a controlled task space of 300 procedurally generated tasks. Models trained on VBVR-Pro show strong transfer beyond the proposed suite across seven external visual reasoning benchmarks such as RISE-Video, MME-CoF-Pro, and BabyVision. 2) Verifiable rewards. VBVR-Pro provides verifiable reward scorers for task-grounded evaluation. Through a systematic study of leading MLLMs as judges, we identify recurring failure modes of the prevalent VLM-as-a-judge paradigm. In contrast, the proposed scorers are grounded in deterministic, task-specific rules, achieve fine-grained alignment with human judgments. Importantly, they serve as reliable reward signals for large-scale multi-task reinforcement learning and demonstrate stronger post-RL performance across visual reasoning tasks. 3) Mechanism study. VBVR-Pro enables controlled modality studies across more than 30 image, video, and interleaved generators. Our analysis shows that video generation remains strongest for tasks requiring persistent spatiotemporal state tracking, while interleaved generation provides a compute-efficient alternative. Critically, ablations and probing suggest the presence of vision-native trajectories that are crucial to visual reasoning. We release all data, models, scorers, and code.

cs.CV↗

arXiv2026

SenseNova-U1.5: Towards Native Unified Visual Intelligence

We launch SenseNova-U1.5, an 8B-MoT native unified multimodal model that understands, reasons about, and generates visual content within an encoder-free and VAE-free architecture. We strengthen its visual interface through spatially coherent patch reconstruction and scale its training with carefully curated generation and editing data, improved task formulation, structural prompt enhancement, and native resolutions of up to 4K. For post-training, we optimize specialized experts for visual aesthetics, bilingual text rendering, infographic generation, and image editing, and consolidate their capabilities through multi-expert on-policy distillation. Across extensive evaluations, SenseNova-U1.5 largely advances image fidelity, text rendering, complex composition, multi-reference editing, and interleaved generation, while improving instruction following and preserving subject identity, geometry, and unmodified regions. Despite limited exposure to structured formats in its generation data, SenseNova-U1.5 generalizes effectively to long, complex, and structured visual instructions, further proving that multimodal understanding can transfer to visual planning and creation. Together, these findings position native unified modelling as a promising path towards systems that perceive, reason and create within a fully end-to-end framework. We will open-source training code, including supervised fine-tuning, reinforcement learning, and on-policy distillation.

cs.CV↗

arXiv2026

ROAM: Robust Organization of Atomic Memories for Agents through Semantic Relations

Long-term language-model agents rely on external memory across interactions. Atomic memories are particularly useful: their fine-grained semantic boundaries enable precise retrieval and direct comparison between observations. Yet accumulating atoms inevitably become redundant, overlapping, or conflicting. Existing methods often ask an LLM manager to add, update, delete, or rewrite memories directly, coupling semantic interpretation, storage decisions, and content generation in one error-prone operation. We introduce ROAM, a relation-guided framework that uses atomicity for management while allowing richer answer-time representations. ROAM classifies incoming--stored atom pairs as independent, equivalent, directionally subsuming, or conflicting, then organizes observations into active Primary and supporting Evidence roles. Fusion subsequently combines complementary details and temporal changes into compact, potentially non-atomic views. Only Primary views are retrieved for answering, preventing redundant or outdated atoms from competing independently. Across models and evaluation settings, ROAM improves answer accuracy by up to 29.8 percentage points. Ablations show complementary benefits from different relations and consistent gains from fusion beyond role organization. Mechanism analysis further finds 15.6-point higher answer-critical source recall and an 11.5-point lower confounder-token share. ROAM remains robust across manager scales.

cs.CL↗

arXiv2026

Entanglement of back-to-back gluon pair

We investigate the polarization correlation of back-to-back gluon pairs in unpolarized $pp$ collisions and compute their density matrix. For gluon pairs produced through $q\bar{q}$ annihilation, the linear polarization correlation is sizable, even reaching unity when the final-state gluons are emitted perpendicular to the beam direction in the $q\bar{q}$ rest frame. Combining with the fact that their helicity remains maximally correlated across different scattering angle, the gluon pair system thus resides in a maximally entangled Bell state in this configuration. In contrast, for the $gg\to gg$ channel, both the helicity and linear polarization correlations are considerably suppressed, producing a separable state. Nonetheless, it still retains an appreciable linear polarization correlation of about $11\%$ at central scattering in unpolarized collisions. The linearly polarization correlation of the back-to-back gluon pair can further increase when the incoming gluons are circularly polarized. Employing the anisotropy of energy correlators, we demonstrate that the linear polarization correlation of gluon pairs can be extracted from the $\cos 2ΔΦ$ modulation.

hep-ph↗

arXiv2026

$\rm S^*(E)$ measurement of the $\rm {}^{12}C({}^{12}C,α){}^{20}Ne$ reaction at astrophysical energies via the Trojan horse method with $\rm ^{16}O$ quasi-free breakup

The 12C(12C,a)20Ne reaction at astrophysical energies is crucial for understanding the carbon burning process in massive star and explosive astrophysical scenarios like Type Ia supernovae and X-ray bursts. However, directly measuring or simply extrapolating its S*(E) factor is extremely challenging due to Coulomb suppression and potential complex resonance structures near the Gamow window (1.5+-0.3 MeV). The THM can circumvent the Coulomb barrier, providing data within the Gamow window without extrapolation. Strong resonances near 1.5 MeV were previously reported by Tumino et al. using THM with 14N=(12C+d), a result that generated significant interest and debate, underscoring the need for further experimental verification. In this work, we selected 16O=(12C+a) as the Trojan-horse nucleus due to its lower binding energy, which favors quasi-free reactions. We performed an indirect measurement of 12C(16O,aa)20Ne at the HI-13 Tandem Accelerator at CIAE. Employing a copper beam-stopper foil, we measured the spectator a-particle within a small angular range around 0, where the quasi-free mechanism predicts its highest concentration. The S*(E) factor of 12C(12C,a)20Ne in the astrophysical energy region was extracted from the measured three-body reaction using THM based on DWBA. Our results confirm the existence of resonances within the Gamow window around 1.5 MeV in both the a0 and a1 channels. Without considering the details of the resonance structures, the overall trend of our results is qualitatively in reasonable agreement with the THM-Tumino2018 and TTIK2025 data, but differs significantly from the trend of the Modified-THM-Muk2019 data. We observe no evidence for hindrance effect in our results.

nucl-ex↗

arXiv2026

Indirect Measurement of the $\rm S^*(E)$ Factor for $\rm {}^{12}C({}^{12}C,\mathit{p}){}^{23}Na$ at Gamow Energies via the Trojan Horse Method with Near-0 degree Spectator Detection

The astrophysical S*(E) factor for the 12C+12C reaction within the Gamow window plays a pivotal role in modeling stellar carbon burning and explosive nucleosynthesis scenarios. However, direct measurements or even simple extrapolations at these energies are severely hindered by Coulomb suppression and the possible presence of narrow resonances. To address this challenge, we performed an indirect measurement of the 12C(16O,ap)23Na reaction at the HI-13 Tandem Accelerator, employing 16O=(12C+a) as the Trojan Horse nucleus. A key innovation of this Trojan Horse Method (THM) study is the implementation of a copper beam-stopper foil, which enabled the detection of spectator particles near 0, the angular region where their yield is maximized under quasi-free kinematics. The S*(E) factor for the 12C(12C,p)23Na reaction in the astrophysically relevant energy range was extracted using the THM formalism based on the DWBA. Our results confirm the presence of resonant structures within the Gamow window around 1.5 MeV in both the p0 and p1 proton channels. No evidence of a hindrance effect is observed in the measured energy range. Without considering the resonance details, the overall trend of our results is qualitatively in reasonable agreement with the THM-Tumino2018 and TTIK2025 data, but differs significantly from the trend of the Modified-THM-Muk2019 data.

nucl-ex↗

arXiv2026

PV-WM: A Heterogeneous Micro-Macro World Model for Articulated Pedestrian-Vehicle Co-Rollout

Local pedestrian-vehicle forecasting spans heterogeneous physical scales: pedestrians combine root locomotion with articulated motion, whereas vehicles are rigid bodies described by kinematic state and oriented extent. Existing road-agent forecasters typically omit pedestrian articulation, while pose forecasters leave vehicle futures outside the learned rollout. We introduce PV-WM, a history-only world model over structured post-perception tracks. It recurrently advances pedestrian root motion, 15-joint articulation, and learned vehicle states within a synchronized heterogeneous state. The generated pedestrian and vehicle chunks supply the next recurrent boundary; vehicle boxes are reconstructed from predicted center and heading with observed extent, and P-V geometry is recomputed after every transition. Relative to a matched one-shot complete-state predictor, recurrent execution reduces Root ADE by 12.7% and MPJPE by 14.8%. Feedback interventions show that later predictions depend on the content, temporal order, and pedestrian identity of generated articulation. Across 824 aligned Waymo contexts, with 797 providing valid future vehicle support, PV-WM reduces Root ADE by 5.2%, MPJPE by 7.6%, P-V distance error by 11.9%, and oriented-box closest-approach error by 5.8% relative to a validation-selected Modular Specialist. The single-network model uses 57.1% fewer parameters, 96.5% lower average FLOPs per local scene, and 25.5% lower measured p95 latency. PV-WM unifies this heterogeneous future state while preserving type-specific pedestrian and vehicle dynamics.

cs.RO↗

arXiv2026

Geometric realizations of representations for $\text{PSL}(2, \mathbb{F}_p)$ and Galois representations arising from defining ideals of modular curves

We construct a geometric realization of representations for $\text{PSL}(2, \mathbb{F}_p)$ by the defining ideals of rational models $\mathcal{L}(X(p))$ of modular curves $X(p)$ over $\mathbb{Q}$, which gives rise to a Rosetta stone for geometric representations of $\text{PSL}(2, \mathbb{F}_p)$. The defining ideal of a modular curve, i.e., an anabelian counterpart of the Eisenstein ideal, is the anabelianization of the Jacobian of this modular curve and is a reification of the fundamental group $π_1$. We show that there exists a correspondence among the defining ideals of modular curves over $\mathbb{Q}$, reducible $\mathbb{Q}(ζ_p)$-rational representations $π_p: \text{PSL}(2, \mathbb{F}_p) \rightarrow \text{Aut}(\mathcal{L}(X(p)))$ of $\text{PSL}(2, \mathbb{F}_p)$, and $\mathbb{Q}(ζ_p)$-rational Galois representations $ρ_p: \text{Gal}(\overline{\mathbb{Q}}/\mathbb{Q}) \rightarrow \text{Aut}(\mathcal{L}(X(p)))$ as well as their modular and surjective realization. It is an anabelian counterpart of the global Langlands correspondence for $\text{GL}(2, \mathbb{Q})$ by the étale cohomology of modular curves as well as an anabelian counterpart of Artin's conjecture, Serre's modularity conjecture, the Fontaine-Mazur conjecture and the $p$-adic local Langlands correspondence for $\text{GL}(2, \mathbb{Q}_p)$. It is an ideal theoretic (i.e. nonlinear) counterpart of Grothendieck's section conjecture and an ideal theoretic (i.e. nonlinear) reification of ``arithmetic theory of $π_1$'' expected by Weil for modular curves. It is also an anabelian counterpart of the theory of Kubert-Lang and Mazur-Wiles on the cuspidal divisor class groups and the Eisenstein ideals of modular curves.

math.NT↗

arXiv2026

Muonium Spectroscopy as a Quantum Sensor for Ultralight Axion Dark Matter

High-intensity muon beams could enable a Muonium-based Axion Search through resonant quantum transitions between Hyperfine states (MASH). Combining theoretical calculations with simulation results, we demonstrate that such a muonium-based experimental approach could complement and tighten constraints on the axion-muon coupling beyond existing limits from the muon $g\!-\!2$ measurement, over the axion mass range of 18--130 $μ$eV. These results establish a new spectroscopic channel in muonium that enables searches for axion and axion-like particle dark matter.

hep-ph↗

arXiv2026

InfraOcc: An Infrastructure Occupancy Benchmark with Static-to-Dynamic Reasoning

Fixed-viewpoint infrastructure sensors repeatedly observe the same traffic space, making roadside 3D occupancy structurally different from ego-vehicle perception: a near-persistent static scaffold is overlaid with sparse, short-lived dynamic events. Existing occupancy benchmarks and methods, however, are built around moving ego vehicles and neither measure nor exploit this structure, instead treating occupancy as flat one-shot voxel classification. We address this gap from both data and model perspectives. We build InfraOcc, to our knowledge, the first real-world infrastructure-side semantic occupancy benchmark, with dense voxel annotations for 290 multi-modal sequences in a fixed roadside frame, a static-dynamic decoupled annotation pipeline, unified camera-only, LiDAR-only, and multi-modal evaluation, and diagnostics for static and dynamic occupancy. InfraOcc shows that static infrastructure fills 97.3% of occupied voxels and persists across frames, whereas dynamic participants have a median occupied-frame ratio of only 1.8% per location, revealing a structural static-dynamic asymmetry beyond semantic long-tailedness. We further propose ProSD-Occ, which reformulates occupancy as progressive static-to-dynamic evidence reasoning: it explains persistent layout, exposes residual dynamic evidence under static-confidence guidance, and recomposes static, dynamic, and free-space evidence into a unified field. ProSD-Occ ranks first in overall, dynamic, static, and geometric occupancy on every track, e.g., a 23.5% relative camera-only dynamic-mIoU gain over the strongest baseline and 65.87 multi-modal overall mIoU, establishing fixed-viewpoint roadside occupancy as a distinct problem with its own reasoning paradigm. The benchmark and code will be publicly available at https://github.com/yanglei18/InfraOcc

cs.CV↗

arXiv2026

GUIDE: Generative Unsupervised Chinese Query Correction via Phonetic and Visual Shared-ID Encoding

Chinese query correction (CQC) is important for search and query recommendation on content platforms, but supervised methods rely on large annotated correction pairs that are costly to maintain as query vocabularies evolve. Unsupervised correction with language models is attractive, yet in the short-query setting, unconstrained generation often over-corrects ambiguous inputs toward high-frequency phrases, causing intent drift. We propose \textsc{GUIDE}, a generative unsupervised framework for CQC based on a confuse-then-clarify paradigm. \textsc{GUIDE} encodes phonetically or visually confusable characters with shared-IDs and reconstructs the original query with an encoder--decoder architecture, which constrains correction to plausible confusion neighborhoods while learning from unlabeled query streams. A time-decayed, query-frequency-weighted objective further supports adaptation to rapidly changing query vocabularies. Experiments on \textit{QSpell 250K} and a large-scale real-world dataset (\textit{KwaiSearch}) show that \textsc{GUIDE} consistently outperforms strong baselines, while online A/B testing further confirms gains in correction quality and downstream engagement.

cs.CL↗

arXiv2026

Dual-Stream Semantic Guidance with Prototype Anchor Calibration for Source-Fully-Free Adaptation of Vision-Language Models

Source-Fully-Free Domain Adaptation (SFF-DA) has emerged as a strategic paradigm to adapt Vision-Language Models (VLMs) without any access to source data or task-specific source models. However, we identify a critical Dual Semantic Drift that hinders this process: static drift arising from the rigidity of fixed class embeddings, and dynamic drift stemming from the divergence of generated captions, causing severe semantic misalignment that intensifies the stability-plasticity dilemma. To address this, we propose DSSG (Dual-Stream Semantic Guidance), an end-to-end framework that reconciles fine-grained plasticity with global stability. Our core contribution is the Dual Semantic Guidance (DSG) module, which integrates a caption stream for domain-specific knowledge with a class-anchor stream to anchor global categorical consistency. Furthermore, a Dynamic Cross-Modal Knowledge Distillation (CMKD) module is introduced to leverage the evolving teacher distribution for calibrating teacher-student consistency. Building upon DSSG, we further introduce Prototype Anchor Calibration (PAC), yielding DSSG-PAC, which periodically calibrates prototype anchors and caches them until the next calibration. This design reduces redundant text-side computation while preserving the adaptability of class guidance to the evolving text space. We further establish SFF-DA risk bounds that relate student risk to semantic-teacher quality and teacher--student discrepancy. Extensive experiments demonstrate that DSSG consistently outperforms current state-of-the-art methods across multiple benchmarks, while DSSG-PAC largely preserves its adaptation performance with 18.9% lower total adaptation time. The code is available at https://github.com/mrmenand/DSSG.

cs.CV↗