arXiv subjects

Jiajun Zhu

Publications and source records attributed to Jiajun Zhu.

At least 19 recordsLinked to original sources

arXiv2026

BehaviorWorldGen: Closing the Loop between Action Models and World Simulators via Controllable Behavior-Aware Structured World Generation

Modern driving action models are increasingly improved in a self-improvement loop, where a learned world simulator imagines future observations and the resulting data is fed back to refine the action model. However, the bottleneck of this loop lies in the simulators' inability to generate behaviorally plausible responses by surrounding agents, making generated data both unrealistic in interaction and imbalanced in distribution. We introduce BehaviorWorldGen, a framework that closes the loop between action models and world simulators through controllable behavior-aware structured world generation. Its core component is BehaviorFlow, a meta-action-conditioned traffic-flow model that injects interpretable behavior controls and jointly generates multi-agent rollouts. BehaviorFlow realizes the specified agent behaviors while allowing surrounding vehicles to respond to the ego and to one another. The resulting rollouts are rendered by a world simulator into realistic multi-view observations, which are paired with corrected interaction-aware trajectories for action-model refinement. Since BehaviorWorldGen uses structured trajectories as the interface between its modules, it is compatible with diverse action models and world simulators. Experiments on world generation, scene extrapolation, and policy refinement demonstrate consistent improvements, with the largest benefits concentrated on difficult interactive scenarios.

cs.RO↗

arXiv2026

GaussianDWM++: Language-Grounded 3D Gaussian Driving World Model for Unified Scene Understanding, Editing, and Multi-Modal Generation

Driving World Models (DWMs) have recently advanced rapidly with generative models, yet most existing methods mainly focus on conditional scene generation and lack explicit 3D scene understanding, language-grounded reasoning, and controllable 4D editing capabilities. Moreover, commonly used point cloud, occupancy, or BEV representations make it difficult to achieve fine-grained alignment between textual information and the underlying 3D scene structure. To address these limitations, we propose a foundation-feature Gaussian driving world model that unifies scene understanding, language-grounded reasoning, controllable 4D editing, and multi-modal generation within a single framework. Specifically, we introduce a foundation-feature Gaussian tokenizer that directly distills Qwen/SigLIP visual-language features into 3D Gaussian primitives, building a compact open-vocabulary Gaussian semantic field. We further design a geometry-aware Gaussian adapter that combines importance-aware hierarchical selection with text-conditioned Perceiver-style cross-attention to aggregate dense Gaussian primitives into compact world tokens. To improve representation compatibility, we introduce a KL-based Gaussian--image distribution alignment objective that aligns Gaussian world tokens with foundation image tokens. Based on the aligned Gaussian representation, our framework further supports instruction-controllable scene editing, including weather-conditioned generation and dynamic vehicle manipulation. Extensive experiments on broader driving benchmarks demonstrate that our method achieves state-of-the-art performance across scene understanding, visual grounding, planning-oriented reasoning, and controllable 4D generation tasks. We will release the code and datasets publicly on Github.

cs.CV↗

arXiv2026

AdaPT: Adaptive Lesson Plan Transformer for Cross-Regional and Differentiated Instruction

Due to educational inequality, high-quality lesson plans often mismatch the needs of disparate educational contexts. Teachers typically modify existing lesson plans to fit new contexts, but current tools instead focus on generating content from scratch, creating additional workload. Moreover, a critical gap remains in supporting teachers to quickly adapt to new learning profiles. To bridge these gaps, we present AdaPT, a system leverages LLMs to support transformation of existing lesson plans for cross-regional and differentiated instruction. AdaPT features an interactive interface that allows teachers to input student profiles, offers structured lesson representation, provides explanations for lesson-plan transformations, automatically adapts lesson content for new contexts, and supports iterative, teacher-in-the-loop refinement. We evaluated AdaPT through a user study with 9 teachers and an expert evaluation with 3 specialists. Results show that AdaPT supports workflows of teachers and offers a promising pathway toward promoting educational equity.

cs.HC↗

arXiv2026

PatchScene: Patch-based Voxel Diffusion for Large-Scale Scene Completion

We propose PatchScene, a novel diffusion-based framework for large-scale LiDAR scene completion. Unlike existing methods that rely on global latent representations or dense voxel grids, PatchScene adopts a patch-based voxel diffusion paradigm that explicitly generates fine-grained geometry within localized 3D regions. To ensure coherent reconstruction at both spatial and temporal scales, we introduce a confidence-guided spatio-temporal fusion mechanism that integrates overlapping patches and adjacent frames in a unified generative process. Furthermore, we design an Annular-Flow diffusion strategy that leverages the radial density pattern of LiDAR scans to progressively propagate high-fidelity information from near-range to far-range regions, enabling spatially unbounded scene completion. Extensive experiments on the SemanticKITTI benchmark demonstrate that PatchScene achieves state-of-the-art performance across all standard metrics, surpassing previous approaches in both geometric accuracy and temporal consistency. Remarkably, the model trained on 20 m LiDAR ranges generalizes effectively to 50 m scenes without retraining, highlighting its strong scalability and generalization capability for real-world autonomous driving applications.

cs.CV↗

arXiv2026

When Do Graph Foundation Models Transfer? A Data-Centric Theory

Graph foundation models (GFMs) aim to reuse a single backbone across diverse graph domains, yet their transfer is often uneven and can exhibit negative transfer. While most prior work improves transfer through architectural or adaptation choices, we ask a data-centric question: which properties of two graph domains determine how much a fixed representation model changes its outputs? Using a graphon-based continuous limit for dense graphs, we show that for both set-based and message-passing tokenizations, any Lipschitz backbone admits an explicit decomposition of cross-domain output shift into (i) graph-specific finite-sample approximation terms and (ii) an intrinsic, relabeling-invariant domain discrepancy capturing structural mismatch. A key ingredient is positional-encoding (PE) stability: we establish stability guarantees for spectral PEs and highlight contrasting behaviors of eigenvector- versus subspace-based PEs. Experiments on synthetic and real graphs validate the theory and translate the decomposition into guidance for data curation in GFM transfer.

cs.LG↗

arXiv2025

ISAC with Affine Frequency Division Multiplexing: An FMCW-Based Signal Processing Perspective

This paper investigates the sensing potential of affine frequency division multiplexing (AFDM) in high-mobility integrated sensing and communication (ISAC) from the perspective of radar waveforms. We introduce an innovative parameter selection criterion that establishes a precise mathematical equivalence between AFDM subcarriers and Nyquist-sampled frequency-modulated continuous-wave (FMCW). This connection not only provides a clear physical insight into AFDM's sensing mechanism but also enables a direct mapping from the DAFT index to delay-Doppler (DD) parameters of wireless channels. Building on this, we develop a novel input-output model in a DD-parameterized DAFT (DD-DAFT) domain for AFDM, which explicitly reveals the inherent DD coupling effect arising from the chirp-channel interaction. Subsequently, we design two matched-filtering sensing algorithms. The first is performed in the time-frequency domain with low complexity, while the second is operated in the DD-DAFT domain to precisely resolve the DD coupling. Simulations show that our algorithms achieve effective pilot-free sensing and demonstrate a fundamental trade-off between sensing performance, communication overhead, and computational complexity. The proposed AFDM outperforms classical AFDM and other variants in most scenarios.

eess.SP↗

arXiv2025

Detect, Investigate, Judge and Determine: A Knowledge-guided Framework for Few-shot Fake News Detection

Few-Shot Fake News Detection (FS-FND) aims to distinguish inaccurate news from real ones in extremely low-resource scenarios. This task has garnered increased attention due to the widespread dissemination and harmful impact of fake news on social media. Large Language Models (LLMs) have demonstrated competitive performance with the help of their rich prior knowledge and excellent in-context learning abilities. However, existing methods face significant limitations, such as the Understanding Ambiguity and Information Scarcity, which significantly undermine the potential of LLMs. To address these shortcomings, we propose a Dual-perspective Knowledge-guided Fake News Detection (DKFND) model, designed to enhance LLMs from both inside and outside perspectives. Specifically, DKFND first identifies the knowledge concepts of each news article through a Detection Module. Subsequently, DKFND creatively designs an Investigation Module to retrieve inside and outside valuable information concerning to the current news, followed by another Judge Module to evaluate the relevance and confidence of them. Finally, a Determination Module further derives two respective predictions and obtain the final result. Extensive experiments on two public datasets show the efficacy of our proposed method, particularly in low-resource settings.

cs.CL↗

arXiv2025

Rethinking Addressing in Language Models via Contexualized Equivariant Positional Encoding

Transformers rely on both content-based and position-based addressing mechanisms to make predictions, but existing positional encoding techniques often diminish the effectiveness of position-based addressing. Many current methods enforce rigid patterns in attention maps, limiting the ability to model long-range dependencies and adapt to diverse tasks. Additionally, most positional encodings are learned as general biases, lacking the specialization required for different instances within a dataset. To address this, we propose con\textbf{T}extualized equivari\textbf{A}nt \textbf{P}osition \textbf{E}ncoding (\textbf{TAPE}), a novel framework that enhances positional embeddings by incorporating sequence content across layers. TAPE introduces dynamic, context-aware positional encodings, overcoming the constraints of traditional fixed patterns. We show that TAPE can provably facilitate LLM reasoning ability by emulating a broader class of algorithms. By enforcing permutation and orthogonal equivariance, TAPE ensures the stability of positional encodings during updates, improving long-context ability. Our method can be easily integrated into pre-trained transformers, offering parameter-efficient fine-tuning with minimal overhead. Extensive experiments show that TAPE achieves superior performance in language modeling, arithmetic reasoning, and long-context retrieval tasks compared to existing positional embedding techniques. Code is available at https://github.com/VITA-Group/TAPE.

cs.CL↗

arXiv2025

ViseGPT: Towards Better Alignment of LLM-generated Data Wrangling Scripts and User Prompts

Large language models (LLMs) enable the rapid generation of data wrangling scripts based on natural language instructions, but these scripts may not fully adhere to user-specified requirements, necessitating careful inspection and iterative refinement. Existing approaches primarily assist users in understanding script logic and spotting potential issues themselves, rather than providing direct validation of correctness. To enhance debugging efficiency and optimize the user experience, we develop ViseGPT, a tool that automatically extracts constraints from user prompts to generate comprehensive test cases for verifying script reliability. The test results are then transformed into a tailored Gantt chart, allowing users to intuitively assess alignment with semantic requirements and iteratively refine their scripts. Our design decisions are informed by a formative study (N=8) that explores user practices and challenges. We further evaluate the effectiveness and usability of ViseGPT through a user study (N=18). Results indicate that ViseGPT significantly improves debugging efficiency for LLM-generated data-wrangling scripts, enhances users' ability to detect and correct issues, and streamlines the workflow experience.

cs.HC↗

arXiv2025

PanoSplatt3R: Leveraging Perspective Pretraining for Generalized Unposed Wide-Baseline Panorama Reconstruction

Wide-baseline panorama reconstruction has emerged as a highly effective and pivotal approach for not only achieving geometric reconstruction of the surrounding 3D environment, but also generating highly realistic and immersive novel views. Although existing methods have shown remarkable performance across various benchmarks, they are predominantly reliant on accurate pose information. In real-world scenarios, the acquisition of precise pose often requires additional computational resources and is highly susceptible to noise. These limitations hinder the broad applicability and practicality of such methods. In this paper, we present PanoSplatt3R, an unposed wide-baseline panorama reconstruction method. We extend and adapt the foundational reconstruction pretrainings from the perspective domain to the panoramic domain, thus enabling powerful generalization capabilities. To ensure a seamless and efficient domain-transfer process, we introduce RoPE rolling that spans rolled coordinates in rotary positional embeddings across different attention heads, maintaining a minimal modification to RoPE's mechanism, while modeling the horizontal periodicity of panorama images. Comprehensive experiments demonstrate that PanoSplatt3R, even in the absence of pose information, significantly outperforms current state-of-the-art methods. This superiority is evident in both the generation of high-quality novel views and the accuracy of depth estimation, thereby showcasing its great potential for practical applications. Project page: https://npucvr.github.io/PanoSplatt3R

cs.CV↗

arXiv2025

Self-Reflective Planning with Knowledge Graphs: Enhancing LLM Reasoning Reliability for Question Answering

Recently, large language models (LLMs) have demonstrated remarkable capabilities in natural language processing tasks, yet they remain prone to hallucinations when reasoning with insufficient internal knowledge. While integrating LLMs with knowledge graphs (KGs) provides access to structured, verifiable information, existing approaches often generate incomplete or factually inconsistent reasoning paths. To this end, we propose Self-Reflective Planning (SRP), a framework that synergizes LLMs with KGs through iterative, reference-guided reasoning. Specifically, given a question and topic entities, SRP first searches for references to guide planning and reflection. In the planning process, it checks initial relations and generates a reasoning path. After retrieving knowledge from KGs through a reasoning path, it implements iterative reflection by judging the retrieval result and editing the reasoning path until the answer is correctly retrieved. Extensive experiments on three public datasets demonstrate that SRP surpasses various strong baselines and further underscore its reliable reasoning ability.

cs.CL↗

arXiv2025

Polarization Decoupling Multi-Port Beam-Splitting Metasurface for Miniaturized Magneto-Optical Trap

In regular magneto-optical trap (MOT) systems, the delivery of six circularly polarized (CP) cooling beams requires complex and bulky optical arrangements including waveplates, mirrors, retroreflectors, etc. To address such technique challenges, we have proposed a beam delivery system for miniaturized MOT entirely based on meta-devices. The key component is a novel multi-port beam-splitting (PD-MPBS) metasurface that relies on both propagation phase and geometric phase. The fabricated samples exhibit high beam-splitting power uniformity (within 4.4%) and polarization purities (91.29%~93.15%). By leveraging such beam-splitting device as well as reflective beam-expanding meta-device, an integrated six-beam delivery system for miniaturized MOT application has been implemented. The experimental results indicate that six expanded beams have been successfully delivered with uniform power (within 9.5%), the desired CP configuration and large overlapping volume (76.2 mm^3). We believe that a miniaturized MOT with the proposed beam delivery system is very promising for portable application of cold atom technology in precision measurement, atomic clock, quantum simulation and computing, etc.

physics.optics↗

arXiv2025

Understanding and Mitigating Bottlenecks of State Space Models through the Lens of Recency and Over-smoothing

Structured State Space Models (SSMs) have emerged as alternatives to transformers. While SSMs are often regarded as effective in capturing long-sequence dependencies, we rigorously demonstrate that they are inherently limited by strong recency bias. Our empirical studies also reveal that this bias impairs the models' ability to recall distant information and introduces robustness issues. Our scaling experiments then discovered that deeper structures in SSMs can facilitate the learning of long contexts. However, subsequent theoretical analysis reveals that as SSMs increase in depth, they exhibit another inevitable tendency toward over-smoothing, e.g., token representations becoming increasingly indistinguishable. This fundamental dilemma between recency and over-smoothing hinders the scalability of existing SSMs. Inspired by our theoretical findings, we propose to polarize two channels of the state transition matrices in SSMs, setting them to zero and one, respectively, simultaneously addressing recency bias and over-smoothing. Experiments demonstrate that our polarization technique consistently enhances the associative recall accuracy of long-range tokens and unlocks SSMs to benefit further from deeper architectures. All source codes are released at https://github.com/VITA-Group/SSM-Bottleneck.

cs.LG↗

arXiv2025

Targetless Intrinsics and Extrinsic Calibration of Multiple LiDARs and Cameras with IMU using Continuous-Time Estimation

Accurate spatiotemporal calibration is a prerequisite for multisensor fusion. However, sensors are typically asynchronous, and there is no overlap between the fields of view of cameras and LiDARs, posing challenges for intrinsic and extrinsic parameter calibration. To address this, we propose a calibration pipeline based on continuous-time and bundle adjustment (BA) capable of simultaneous intrinsic and extrinsic calibration (6 DOF transformation and time offset). We do not require overlapping fields of view or any calibration board. Firstly, we establish data associations between cameras using Structure from Motion (SFM) and perform self-calibration of camera intrinsics. Then, we establish data associations between LiDARs through adaptive voxel map construction, optimizing for extrinsic calibration within the map. Finally, by matching features between the intensity projection of LiDAR maps and camera images, we conduct joint optimization for intrinsic and extrinsic parameters. This pipeline functions in texture-rich structured environments, allowing simultaneous calibration of any number of cameras and LiDARs without the need for intricate sensor synchronization triggers. Experimental results demonstrate our method's ability to fulfill co-visibility and motion constraints between sensors without accumulating errors.

cs.RO↗

arXiv2024

Towards Understanding Sensitive and Decisive Patterns in Explainable AI: A Case Study of Model Interpretation in Geometric Deep Learning

The interpretability of machine learning models has gained increasing attention, particularly in scientific domains where high precision and accountability are crucial. This research focuses on distinguishing between two critical data patterns -- sensitive patterns (model-related) and decisive patterns (task-related) -- which are commonly used as model interpretations but often lead to confusion. Specifically, this study compares the effectiveness of two main streams of interpretation methods: post-hoc methods and self-interpretable methods, in detecting these patterns. Recently, geometric deep learning (GDL) has shown superior predictive performance in various scientific applications, creating an urgent need for principled interpretation methods. Therefore, we conduct our study using several representative GDL applications as case studies. We evaluate thirteen interpretation methods applied to three major GDL backbone models, using four scientific datasets to assess how well these methods identify sensitive and decisive patterns. Our findings indicate that post-hoc methods tend to provide interpretations better aligned with sensitive patterns, whereas certain self-interpretable methods exhibit strong and stable performance in detecting decisive patterns. Additionally, our study offers valuable insights into improving the reliability of these interpretation methods. For example, ensembling post-hoc interpretations from multiple models trained on the same task can effectively uncover the task's decisive patterns.

cs.LG↗

arXiv2024

Retrosynthesis Prediction via Search in (Hyper) Graph

Predicting reactants from a specified core product stands as a fundamental challenge within organic synthesis, termed retrosynthesis prediction. Recently, semi-template-based methods and graph-edits-based methods have achieved good performance in terms of both interpretability and accuracy. However, due to their mechanisms these methods cannot predict complex reactions, e.g., reactions with multiple reaction center or attaching the same leaving group to more than one atom. In this study we propose a semi-template-based method, the \textbf{Retro}synthesis via \textbf{S}earch \textbf{i}n (Hyper) \textbf{G}raph (RetroSiG) framework to alleviate these limitations. In the proposed method, we turn the reaction center identification and the leaving group completion tasks as tasks of searching in the product molecular graph and leaving group hypergraph respectively. As a semi-template-based method RetroSiG has several advantages. First, RetroSiG is able to handle the complex reactions mentioned above by its novel search mechanism. Second, RetroSiG naturally exploits the hypergraph to model the implicit dependencies between leaving groups. Third, RetroSiG makes full use of the prior, i.e., one-hop constraint. It reduces the search space and enhances overall performance. Comprehensive experiments demonstrated that RetroSiG achieved competitive results. Furthermore, we conducted experiments to show the capability of RetroSiG in predicting complex reactions. Ablation experiments verified the efficacy of specific elements, such as the one-hop constraint and the leaving group hypergraph.

q-bio.QM↗

arXiv2024

Theoretical Design of Mono-Elemental Ferroelectricity with Tunable Spin Textures in Bilayer Tellurium

2D Ferroelectricity with switchable electric polarization has drawn widespread attention in condensed matter physics due to its crucial applications in non-volatile memory and ferroelectric spin devices. Despite recent progress in 2D ferroelectric, achieving the mono-elemental ferroelectricity still remains a great challenge because most nonmetallic mono-elemental materials are stabilized in nonpolar crystal structures. In this work, we theoretically designed mono-elemental ferroelectricity with tunable and significant spin textures in bilayer tellurium (BL-Te). Comprehensive quantitative polarization calculations demonstrate that asymmetric stacking in BL-Te can generate out-of-plane (OOP) polarization with a magnitude of 0.78 pC/m. This polarization stems from distinguishing interlayer and intra-layer contributions. Moreover, these stacked BL-Te, characterized by significant spin-orbit coupling, serve as an ideal platform for investigating both conventional spin polarization and layer-dependent/hidden spin polarization through ferroelectric reversion. Our work not only broaden the category of 2D mono-elemental ferroelectric but also offer a new platform for multifunctional nanodevices.

cond-mat.mtrl-sci↗

arXiv2023

A Low-Complexity Range Estimation with Adjusted Affine Frequency Division Multiplexing Waveform

Affine frequency division multiplexing (AFDM) is a recently proposed communication waveform for time-varying channel scenarios. As a chirp-based multicarrier modulation technique it can not only satisfy the needs of multiple scenarios in future mobile communication networks but also achieve good performance in radar sensing by adjusting the built-in parameters, making it a promising air interface waveform in integrated sensing and communication (ISAC) applications. In this paper, we investigate an AFDM-based radar system and analyze the radar ambiguity function of AFDM with different built-in parameters, based on which we find an AFDM waveform with the specific parameter c2 owns the near-optimal time-domain ambiguity function. Then a low-complexity algorithm based on matched filtering for high-resolution target range estimation is proposed for this specific AFDM waveform. Through simulation and analysis, the specific AFDM waveform has near-optimal range estimation performance with the proposed low-complexity algorithm while having the same bit error rate (BER) performance as orthogonal time frequency space (OTFS) using simple linear minimum mean square error (LMMSE) equalizer.

eess.SP↗