arXiv subjects

Tianran Liu

Publications and source records attributed to Tianran Liu.

9 recordsLinked to original sources

arXiv2026

OccSim: Multi-kilometer Simulation with Long-horizon Occupancy World Models

Data-driven autonomous driving simulation has long been constrained by its heavy reliance on pre-recorded driving logs or spatial priors, such as HD maps. This fundamental dependency severely limits scalability, restricting open-ended generation capabilities to the finite scale of existing collected datasets. To break this bottleneck, we present OccSim, the first occupancy world model-driven 3D simulator. OccSim obviates the requirement for continuous logs or HD maps; conditioned only on a single initial frame and a sequence of future ego-actions, it can stably generate over 3,000 continuous frames, enabling the continuous construction of large-scale 3D occupancy maps spanning over 4 kilometers for simulation. This represents an >80x improvement in stable generation length over previous state-of-the-art occupancy world models. OccSim is powered by two modules: W-DiT based static occupancy world model and the Layout Generator. W-DiT handles the ultra-long-horizon generation of static environments by explicitly introducing known rigid transformations in architecture design, while the Layout Generator populates the dynamic foreground with reactive agents based on the synthesized road topology. With these designs, OccSim can synthesize massive, diverse simulation streams. Extensive experiments demonstrate its downstream utility: data collected directly from OccSim can pre-train 4D semantic occupancy forecasting models to achieve up to 67% zero-shot performance on unseen data, outperforming previous asset-based simulator by 11%. When scaling the OccSim dataset to 5x the size, the zero-shot performance increases to about 74%, while the improvement over asset-based simulators expands to 22.1%.

cs.CV↗

arXiv2026

AutoWorld: Learning Multi-Agent Traffic Simulation with Self-Supervised World Models

Simulation with realistic traffic agents is essential for validating autonomous driving systems. Existing data-driven simulators learn agent behavior from higher-level abstractions such as 3D bounding boxes and polylines, inferred by upstream perception pipelines. These lossy abstractions discard sensory context that directly shapes agent behavior, limiting the distributional realism that simulation aims to reproduce. To address this limitation, we propose AutoWorld, a traffic simulation framework that grounds agent behavior in raw sensor observations through a self-supervised world model trained on LiDAR occupancy data. Given world model samples, AutoWorld constructs a coarse-to-fine predictive scene context as input to a multi-agent motion generation model. Furthermore, we designed a motion-aware latent supervision objective that enriches AutoWorld's latent representation of scene dynamics. To better exploit this latent space during inference, AutoWorld employs a cascaded Determinantal Point Process framework to guide diversity-aware sampling across both the world model and motion model. Experiments on the Waymo Sim Agents Challenge (WOSAC) demonstrate that AutoWorld achieves competitive performance, with larger gains in partially-observed scenarios where trajectory abstractions are most limited. We further show that grounding simulation in raw LiDAR through AutoWorld scales better with additional data than trajectory-only and LiDAR-conditioning baselines. Ablations confirm the contribution of each component.

cs.RO↗

arXiv2025

Towards foundational LiDAR world models with efficient latent flow matching

LiDAR-based world models offer more structured and geometry-aware representations than their image-based counterparts. However, existing LiDAR world models are narrowly trained; each model excels only in the domain for which it was built. Can we develop LiDAR world models that exhibit strong transferability across multiple domains? We conduct the first systematic domain transfer study across three demanding scenarios: (i) outdoor to indoor generalization, (ii) sparse-beam & dense-beam adaptation, and (iii) non-semantic to semantic transfer. Given different amounts of fine-tuning data, our experiments show that a single pre-trained model can achieve up to 11% absolute improvement (83% relative) over training from scratch and outperforms training from scratch in 30/36 of our comparisons. This transferability of dynamic learning significantly reduces the reliance on manually annotated data for semantic occupancy forecasting: our method exceed the previous semantic occupancy forecasting models with only 5% of the labeled training data required by prior models. We also observed inefficiencies of current LiDAR world models, mainly through their under-compression of LiDAR data and inefficient training objectives. To address this, we propose a latent conditional flow matching (CFM)-based frameworks that achieves state-of-the-art reconstruction accuracy using only half the training data and a compression ratio 6 times higher than that of prior methods. Our model achieves SOTA performance on future-trajectory-conditioned semantic occupancy forecasting while being 23x more computationally efficient (a 28x FPS speedup); and achieves SOTA performance on semantic occupancy forecasting while being 2x more computationally efficient (a 1.1x FPS speedup).

cs.CV↗

arXiv2025

A closed-loop AI framework for hypothesis-driven and interpretable materials design

Scientific hypothesis generation is central to materials discovery, yet current approaches often emphasize either conceptual (idea-to-data) reasoning or data-driven (data-to-idea) analysis, rarely achieving an effective integration of both. Here, we present a generalizable active learning workflow that integrates top-down, theory-driven hypothesis generation, guided by a large language model. This is complemented by bottom-up, data-driven hypothesis testing through a root-cause association study. We demonstrate this approach through the design of equimolar quinary-cation two-dimensional perovskite, a chemically complex system with over 850,000 possible cation combinations. In the top-down component, the large language model drives closed-loop optimization by proposing candidates that are likely to achieve phase purity, leveraging domain knowledge and chain-of-thought reasoning. With each iteration, the model identifies an increasing number of near phase-pure compositions, sampling less than 0.004% of the design space. In parallel, the bottom-up association study identifies molecular features with statistically significant influences on phase purity. The integration of these approaches enables the convergence of conceptual and statistical hypotheses, leading to generalizable and rational design rules for phase-pure quinary-cation two-dimensional perovskites. As a proof of concept, we applied the optimized phase-pure quinary-cation two-dimensional perovskite film as a surface capping layer in perovskite solar cells, achieving good performance and stability. Our framework enables the development of interpretable and generalizable design rules that are applicable to a wide range of optimization processes within complex design spaces, providing a foundational strategy for rational, scalable, and efficient materials discovery.

cond-mat.mtrl-sci↗

arXiv2025

Disentangling the Effects of Simultaneous Environmental Variables on Perovskite Synthesis and Device Performance via Interpretable Machine Learning

Despite the rapid rise in perovskite solar cell efficiency, poor reproducibility remains a major barrier to commercialization. Film crystallization and device performance are highly sensitive to environmental factors during fabrication, yet their complex interactions are not well understood. In this work, we present a systematic framework to investigate the influence of both individual and coupled environmental variables on device efficiency and crystallization kinetics. We developed an integrated fabrication platform with precise, independent control over ambient solvent partial pressure, absolute humidity, and temperature during spin-coating and thermal-annealing processes, respectively. Using the platform, we implemented a closed-loop Bayesian optimization framework to efficiently explore the multi-dimensional processing space. We mapped the impact of these environmental variables on device performance and identified coupled effects among them. In-situ grazing-incidence wide-angle X-ray scattering measurements further validated these findings by revealing a nonlinear interaction between absolute humidity and solvent partial pressure during spin-coating, which affects crystallization dynamics. To isolate and quantify these interactions, we developed an interpretable machine learning approach that combines knowledge distillation with Shapley interaction analysis. The model revealed the contribution of each interaction varies across different processing conditions. Our study highlights the importance of integrated ambient sensing and control to achieve repeatable perovskite solar cells, and demonstrates the utility of combining active learning with interpretable machine learning to navigate complex, high-dimensional processing landscapes.

cond-mat.mtrl-sci↗

arXiv2025

Multimodal transformers with elemental priors for phase classification of X-ray diffraction spectra

Classifying a crystalline solid's phase using X-ray diffraction (XRD) is a challenging endeavor, first because this is a poorly constrained problem as there are nearly limitless candidate phases to compare against a given experimental spectrum, and second because experimental signals are confounded by overlapping peaks, preferred orientations, and phase mixtures. To address this challenge, we develop Chem-XRD, a multimodal framework based on vision transformer (ViT) architecture with two defining features: (i) crystallographic data is constrained by elemental priors and (ii) phase is classified according to probabilities (not absolutes). Elemental information can be extracted from pre-synthesis precursors or post-synthesis elemental analysis. By combining structural and elemental modalities, Chem-XRD simultaneously predicts both the number and identity of phases of lead-halide perovskite materials and their decomposition products. Through integrated gradient calculations, we show that the model can adeptly adjust the contributions of structural and elemental modalities toward the final prediction of phase identification, achieving a level of interpretability beyond what self-attention mechanisms can achieve.

physics.app-ph↗

arXiv2024

FSMDet: Vision-guided feature diffusion for fully sparse 3D detector

Fully sparse 3D detection has attracted an increasing interest in the recent years. However, the sparsity of the features in these frameworks challenges the generation of proposals because of the limited diffusion process. In addition, the quest for efficiency has led to only few work on vision-assisted fully sparse models. In this paper, we propose FSMDet (Fully Sparse Multi-modal Detection), which use visual information to guide the LiDAR feature diffusion process while still maintaining the efficiency of the pipeline. Specifically, most of fully sparse works focus on complex customized center fusion diffusion/regression operators. However, we observed that if the adequate object completion is performed, even the simplest interpolation operator leads to satisfactory results. Inspired by this observation, we split the vision-guided diffusion process into two modules: a Shape Recover Layer (SRLayer) and a Self Diffusion Layer (SDLayer). The former uses RGB information to recover the shape of the visible part of an object, and the latter uses a visual prior to further spread the features to the center region. Experiments demonstrate that our approach successfully improves the performance of previous fully sparse models that use LiDAR only and reaches SOTA performance in multimodal models. At the same time, thanks to the sparse architecture, our method can be up to 5 times more efficient than previous SOTA methods in the inference process.

cs.CV↗

arXiv2023

What You See Is What You Detect: Towards better Object Densification in 3D detection

Recent works have demonstrated the importance of object completion in 3D Perception from Lidar signal. Several methods have been proposed in which modules were used to densify the point clouds produced by laser scanners, leading to better recall and more accurate results. Pursuing in that direction, we present, in this work, a counter-intuitive perspective: the widely-used full-shape completion approach actually leads to a higher error-upper bound especially for far away objects and small objects like pedestrians. Based on this observation, we introduce a visible part completion method that requires only 11.3\% of the prediction points that previous methods generate. To recover the dense representation, we propose a mesh-deformation-based method to augment the point set associated with visible foreground objects. Considering that our approach focuses only on the visible part of the foreground objects to achieve accurate 3D detection, we named our method What You See Is What You Detect (WYSIWYD). Our proposed method is thus a detector-independent model that consists of 2 parts: an Intra-Frustum Segmentation Transformer (IFST) and a Mesh Depth Completion Network(MDCNet) that predicts the foreground depth from mesh deformation. This way, our model does not require the time-consuming full-depth completion task used by most pseudo-lidar-based methods. Our experimental evaluation shows that our approach can provide up to 12.2\% performance improvements over most of the public baseline models on the KITTI and NuScenes dataset bringing the state-of-the-art to a new level. The codes will be available at \textcolor[RGB]{0,0,255}{\url{https://github.com/Orbis36/WYSIWYD}

cs.CV↗

arXiv2022

Testing Pre-trained Language Models' Understanding of Distributivity via Causal Mediation Analysis

To what extent do pre-trained language models grasp semantic knowledge regarding the phenomenon of distributivity? In this paper, we introduce DistNLI, a new diagnostic dataset for natural language inference that targets the semantic difference arising from distributivity, and employ the causal mediation analysis framework to quantify the model behavior and explore the underlying mechanism in this semantically-related task. We find that the extent of models' understanding is associated with model size and vocabulary size. We also provide insights into how models encode such high-level semantic knowledge.

cs.CL↗