ECCV2026 Autonomous Driving AI paper notes paper summaries 3D Object Detection Adversarial Robustness Agents Multimodal/VLM Diffusion Models

🚗 Autonomous Driving¶

🎞️ ECCV2026 · 159 paper notes

📌 Same area in other venues: 🧠 NeurIPS2026 (1) · 📷 CVPR2026 (157) · 🔬 ICLR2026 (50) · 🧪 ICML2026 (8) · 🤖 AAAI2026 (56) · 🧠 NeurIPS2025 (47)

🔥 Top topics: Autonomous Driving ×30 · 3D Object Detection ×13 · Adversarial Robustness ×11 · Agents ×8 · Multimodal/VLM ×7

Masked Depth Modeling for Spatial Perception: MDM turns missing sensor depth into physically motivated masks and learns geometric representations through joint attention over complete RGB and remaining depth, reducing NYUv2 Extreme RMSE from PromptDA's 0.324 to 0.167 while improving real-world robotic grasping.
MindDrive: A Vision-Language-Action Model for Autonomous Driving via Online Reinforcement Learning: MindDrive uses a shared language model with separate LoRA adapters for decision and action experts, restricting online exploration to discrete driving intentions and optimizing decisions through rewards from executed trajectories; its 3B version reaches 80.59 DS and 58.26% SR on Bench2Drive, although validation remains limited to CARLA simulation.
UniDrive-WM: Unified Understanding, Planning and Generation World Model for Autonomous Driving: UniDrive-WM jointly achieves scene understanding, trajectory planning, and trajectory-conditioned future image generation within a single VLM framework. Through a differentiable bidirectional coupling of planning and generation, it improves the Driving Score from 77.74 to 79.31 in Bench2Drive closed-loop evaluation, reduces the L2 trajectory error by 7.3%, and decreases the collision rate by 10.4%. Furthermore, it systematically analyzes the speed-fidelity trade-off between two complementary image generation branches (discrete autoregressive and continuous diffusion).
DriveVA: Video Action Models are Zero-Shot Drivers: DriveVA proposes a unified video-action world model that jointly decodes future video prediction and trajectory planning within a single DiT diffusion generation process. Leveraging the spatio-temporal priors of a large-scale video generation model (Wan2.2-TI2V-5B), it achieves strong zero-shot cross-domain generalization, obtaining 90.9 PDMS on NAVSIM, and reducing the average L2 error and collision rate by 78.9%/52.5% and 83.3%/52.4% on nuScenes and Bench2Drive, respectively.
SparseDriveV2: Scoring is All You Need for End-to-End Autonomous Driving: SparseDriveV2 replaces dynamic trajectory generation with a super-dense static space composed of paths and velocity profiles, selecting actions through factor-level filtering and trajectory-level scoring to reach 92.0 PDMS on NAVSIM v1 and 90.1 EPDMS under the corrected NAVSIM v2 protocol with ResNet-34.
Driver-WM: A Driver-Centric Traffic-Conditioned Latent World Model for In-Cabin Dynamics Rollout: Modeling "how the in-cabin driver will move in the future" as a rollout problem in a latent space, causally driven by out-of-cabin traffic events: utilizing frozen Qwen3-VL features as the perception interface, a dual-stream architecture separately encodes traffic and the driver. A "gated causal injection" mechanism allows the external environment to unidirectionally modulate the autoregressive evolution of the driver's internal state. This is ultimately decoded into the human skeleton trajectory for the next 5 steps (augmented by behavioral/emotional semantics as auxiliary regularization), achieving robust long-term prediction on safety-critical, high-motion sequences in the AIDE dataset.
UniDriveDreamer: A Single-Stage Multimodal World Model for Autonomous Driving: UniDriveDreamer calibrates video and LiDAR latents before feeding them into one generative Transformer, jointly synthesizing multisensor sequences through bidirectional interaction and achieving FID 2.81 and FVD 11.44 on nuScenes, with augmentation gains in a controlled detection experiment.
ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving: A unified understanding-generation framework, ExploreVLA, is proposed. It introduces future RGB + depth image generation as a dense world modeling objective within the VLA driving model. The image prediction uncertainty of the world model is leveraged as an intrinsic exploration reward. Under a safety-gated GRPO, the policy is enabled to surpass the imitation learning distribution and discover diverse yet safe driving behaviors, achieving state-of-the-art results of \(93.7\) PDMS and \(88.8\) EPDMS on NAVSIM.
Beyond Imitation: Learning Safe End-to-End Autonomous Driving from Hard Negatives: BeyondDrive trains a planner to approach expert trajectories while avoiding generated trajectories that are nearby but unsafe, improving optimized LTF from 88.7 to 89.7 PDMS on NAVSIM v1 and from 89.1 to 90.1 EPDMS on NAVSIM v2, with additional evidence from other planners and zero-shot HUGSIM evaluation.
CausalDrive: Real-time Causal World Models for Autonomous Driving: CausalDrive generates reactive driving videos from only an initial front-view image, an ego trajectory, and a global interaction prompt; a causal autoregressive teacher and context-forced DMD reduce inference to 1-4 steps, reaching 12.4 FPS on one A100 and an 82.0% yielding rate under the Polite condition on SocioDrive-Bench, without eliminating all false collisions.
HAD: Combining Hierarchical Diffusion with Metric-Decoupled RL for End-to-End Driving: HAD filters driving intentions before generating and refining structured local trajectories, trains with per-metric reinforcement learning and offline reward retrieval, and reaches 88.6 EPDMS on NAVSIM v2; its camera-only HAD-L variant reaches 47.5 Route Completion on the HUGSIM public set.
AutoWeather4D: Autonomous Driving Video Weather Conversion via G-Buffer Dual-Pass Editing: AutoWeather4D extracts metric geometry and materials from driving videos, edits weather interactions and light transport separately, and uses constrained video diffusion to improve sensor realism, reaching a 0.2586 CLIP score and 0.915 vehicle detection IoU on Waymo; its main advantage is spatially grounded, parametric weather and lighting control rather than a large numerical lead over generative baselines.
BEV-GS: Feed-forward Gaussian Splatting in Bird's-Eye-View for Road Reconstruction: BEV-GS replaces perspective-pixel Gaussian prediction with elevation and texture prediction on a regular BEV grid, achieving 1.73 cm average absolute elevation error, 28.36 dB novel-view PSNR, and 26.3 FPS parameter prediction on RSRD; single-frame refers to inference input, while training still uses elevation labels and stereo-image supervision.
Composing Driving Worlds through Disentangled Control for Adversarial Scenario Generation: CompoSIA injects scene structure, element identity, and ego action through different control pathways in a video diffusion model, enabling compositional driving edits with an identity-control FVD of 149.15 and exposing higher planning collision rates under edited scenarios in an open-loop nuScenes evaluation with Epona.
LangDriveCTRL: Natural Language Controllable Driving Scene Editing with Multi-modal Agents: LangDriveCTRL reconstructs each real driving video into a 3DGS scene graph of a static background plus dynamic object nodes with trajectories, then lets an LLM orchestrator dispatch multimodal object grounding, node editing, behavior editing, behavior review, video diffusion and video review agents and tools, turning natural-language instructions into object-level removal/insertion/replacement and multi-object behavior editing with nearly twice the instruction alignment of prior state of the art, while preserving photorealism, scene structure and traffic realism.
Optimization-Guided Diffusion for Interactive Scene Generation: Without retraining or modifying any diffusion-based scene generator, OMEGA re-anchors the model-predicted clean sample inside every reverse denoising step through a constrained optimization within a KL-bounded trust region, and schedules structural versus interaction guidance via a two-phase (Warm-up / Rolling-Zero) denoising scheme — its adversarial variant ΩAdv further casts ego–attacker interaction as a distributional game approximated by sensitivity-enhanced iterative best response. It lifts the scene validity rate from 32.35% to 72.27% on nuPlan free exploration, from 61.71% to 81.95% on zero-shot Waymo, and from 11% to 80% under goal-conditioned controllability, while generating roughly 5× more near-collision frames with TTC below 3 s without sacrificing realism.
CritiqueDriveVLM: From Verifier-Guided Reinforcement Learning to Latent Thought Distillation for Autonomous Driving: CritiqueDriveVLM trains a self-correcting teacher using verifier scores and textual critiques, then aligns the teacher's reasoning-end state with the student's answer-start state, enabling CoT-free inference at 68.59% MCQ and reducing mean latency from 3482 ms to 416 ms, without preserving the teacher's 76.54% accuracy.
Deconfounded Lifelong Learning for Autonomous Driving via Dynamic Knowledge Spaces: DeLL introduces lifelong learning to closed-loop end-to-end autonomous driving for the first time. It utilizes a Dirichlet Process Mixture Model (DPMM) to construct dual dynamic knowledge spaces (explicit and implicit) and uses dynamically growing knowledge anchors as mediator variables to achieve deconfounding via front-door adjustment, achieving superior lifelong learning performance over PackNet and ER on Bench2Drive with zero extra storage overhead.
LeAD-M3D: Leveraging Asymmetric Distillation for Real-Time Monocular 3D Detection: LeAD-M3D integrates three core components on top of an image-only YOLOv10-M3D baseline: Asymmetric MixUp Denoising Distillation (A2D2), a 3D-aware Consistent Matcher (CM3D) that incorporates 3D overlap into matching scores, and Confidence Gated Inference (CGI3D) that runs the 3D regression head only on high-confidence regions. Without relying on LiDAR, stereo images, or geometric priors, it achieves state-of-the-art accuracy on KITTI, Waymo, and Rope3D, while running up to 3.6 times faster than models of comparable accuracy and exceeding 60 FPS under TensorRT for its largest variant.
LiFlow: Flow Matching for 3D LiDAR Scene Completion: LiFlow transports a locally perturbed single LiDAR scan toward a complete scene, using nearest neighbor flow matching for motion supervision and Chamfer distance matching for coverage, achieving low geometric error with few integration steps while remaining weaker on fine-grained occupancy.
RadarGen: Automotive Radar Point Cloud Generation from Cameras: RadarGen extracts BEV visual conditions from surround cameras at two timestamps, jointly generates radar density, RCS, and Doppler maps, and recovers sparse points, improving foreground box hit rate from 0.37 to 0.66 on MAN TruckScenes while remaining substantially below real radar in downstream detection quality.
WorldFlow3D: Flowing Through 3D Distributions for Unbounded World Generation: WorldFlow3D reformulates 3D scene generation as flow-matching transport across a hierarchy of data distributions — noise → coarse geometry → fine geometry → textured appearance — all directly in TUDF voxel space with no latent autoencoder, and achieves theoretically unbounded synthesis at inference through feather-weighted velocity averaging over overlapping chunks; on Waymo driving and 3D-FRONT indoor data it beats XCube, BlockFusion, WorldGrow, and LT3SD on essentially every metric except one unconditional JSD cell, trains at least 2× faster than latent-space methods, and tops the user study with a Bradley-Terry score of 0.692 and an 88% win rate.
YouTube-Occ: Learning Indoor 3D Semantic Occupancy Prediction from YouTube Videos: YouTube-Occ converts real-estate videos into indoor geometric and semantic pseudo-labels, then pretrains occupancy networks through differentiable rendering and dual-alignment feature distillation; with only 10% of NYUv2 labels for fine-tuning, web-video pretraining improves Symphonies from 17.43 to 22.52 mIoU (+5.09), whereas direct pseudo-label supervision yields 37.38 rather than 40.17 mIoU when the web dataset replaces NYUv2 in joint training with Occ-ScanNet.
A Physics-Grounded Benchmark for Multi-Agent Dynamics in World Models: CrashTwin combines traffic-interaction data, calibration-free reconstruction, and separate physical diagnostics to show that visually coherent world-model videos can still contain incorrect momentum exchange and kinetic-energy increases; its contribution is an evaluation framework, not a new video generator.
ASSCG: Just-Right Gating over Chattering for Fast–Slow LLM Planning in Autonomous Driving: ASSCG learns not only when to query a slow language model but also when to reuse or suppress its cached guidance, using temporal gating and compute-aware reinforcement learning to raise AsyncDriver's nuPlan Hard20 score from 65.00 to 67.28 while reducing average end-to-end inference latency from 0.80 to 0.32 seconds per frame.
BeyondSight: Object Permanence for End-to-End Autonomous Driving: BeyondSight separates historical actor propagation from current-image updates in SparseDrive and trains with annotations spanning unobservable intervals, keeping fully occluded actors available to prediction and planning while raising unobservable-actor mAP from 0 to 0.249 and reducing standard nuScenes average open-loop planning L2 from 0.61 to 0.54.
Driving like yourself: A Benchmark for Closed-Loop Personalized End-to-End Autonomous Driving: Person2Drive connects human-in-the-loop simulator collection, individual style-distribution evaluation, and style-guided trajectory-head fine-tuning, raising average individual style similarity across two routes from 0.6652 to 0.6800, although driving-competence evidence remains simulation-based and not every metric improves.
EgoDyn-Bench: Evaluating Ego-Motion Understanding in Vision-Centric Foundation Models for Autonomous Driving: EgoDyn-Bench uses a deterministic oracle to map continuous ego-vehicle kinematics onto discrete motion concepts, casting ego-motion understanding as a semantic VideoQA benchmark over 3-second clips; a large-scale audit of 20+ vision-centric foundation models (closed-source MLLMs, open-source VLMs, driving VLAs) exposes a Perception Bottleneck — models possess physically coherent reasoning yet cannot ground it in visual observations, frequently losing to classical geometric baselines, while supplying explicit trajectory text restores performance and reveals that ego-motion logic is drawn almost exclusively from the language modality.
LaGen: Towards Autoregressive LiDAR Scene Generation: LaGen introduces the autoregressive framework to LiDAR scene generation for the first time. Starting from a single-frame point cloud and conditioning on 3D bounding boxes and ego-vehicle states, it generates high-fidelity, long-sequence LiDAR scenes frame-by-frame, supporting interactive closed-loop simulation.
Learning Ego-Centric BEV Representations from a Perspective-Privileged View: Cross-View Supervision for Online HD Map Construction: Treating ego-aligned aerial imagery as a "perspective-privileged view" available only at training time, a frozen aerial encoder supplies dense BEV feature-level alignment supervision to a camera BEV encoder (stabilized by channel normalization and an affine adapter), lifting StreamMapNet on nuScenes/AID4AD from 34.1 to 38.0 mAP (60×30 m) and from 22.4 to 32.3 mAP (100×50 m, +44% relative) without changing the inference architecture.
LiDAR-EVS: Enhance Extrapolated View Synthesis for 3D Gaussian Splatting with Pseudo-LiDAR Supervision: Without collecting any additional multi-pass data, single-traversal LiDAR sequences are turned into pseudo-LiDAR supervision for poses outside the training trajectory (multi-frame static fusion, extrapolated pose transformation, occlusion curling, incidence-angle intensity correction), and a dropout regularizer partitioned by the sensor's elevation/distance geometry is added; on the real extrapolations of Para-Lane this cuts Chamfer Distance from 0.35 m to 0.18 m and depth error from 0.003 m² to 0.001 m² while leaving interpolation and camera rendering quality intact.
LiSTAR: Ray-Centric World Models for 4D LiDAR Sequences in Autonomous Driving: LiSTAR voxelizes LiDAR point clouds into a Hybrid-Cylindrical-Spherical (HCS) 3D grid aligned with the sampling geometry of a spinning sensor, encodes spatio-temporal features with a START module that attends along sensor rays, and generates with a MaskSTART masked model, beating Copilot4D and OpenDWM on nuScenes/KITTI reconstruction, prediction, and controllable generation — up to 85% relative IoU gain in reconstruction, up to 64% lower prediction L1 Med, and 61% lower generation MMD.
LMGenDrive: Bridging Multimodal Understanding and Generative World Modeling for End-to-End Driving: LMGenDrive stitches a LLaMA-family LLM and a multi-view diffusion world model into one model through two sets of queries: the world queries carry scene evolution and serve as the conditioning signal for video generation, and the generation loss back-propagates through them into the LLM, so that "imagining the future" becomes a supervision signal that improves "understanding and planning" — lifting DS from 48.9 to 62.2 on the CARLA LangAuto closed-loop benchmark.
Long-term Traffic Simulation via Structured Autoregressive Modeling: RosettaSim models long-term multi-agent traffic simulation as a structured autoregressive sequence generation problem. By utilizing frozen/partially frozen LLMs (Qwen2.5-0.5B) as structural priors, it projects scene topology, agent states, and generation intents into variable-length token sequences to uniformly process motion prediction and agent population dynamics. Meanwhile, it proposes a Retrieval-based Traffic Evaluation (RTE) framework, using semantically similar real-world scenes as reference anchors instead of global distribution matching, achieving SOTA on both short-term and long-term simulation on WOSAC.
MobileOcc: A Human-Aware Semantic Occupancy Dataset for Mobile Robots: Built on synchronized images and LiDAR from the UT Campus Object Dataset (CODa), MobileOcc is a semantic occupancy dataset for mobile robots in pedestrian-dense environments: deformable human meshes obtained by jointly optimizing a monocular SMPL initialization against LiDAR points replace rigid bounding-box proxies, covering 116,511 frames at up to 0.02 m label resolution, together with a benchmark suite for occupancy prediction and pedestrian velocity prediction. Static voxel labels reach 71.2 mIoU and pedestrian voxels reach 92.5 F1 against a LiDAR body-surface reference, while existing occupancy models attain only 26.6-29.9 mIoU on the benchmark.
PLOT: Pseudo-Labeling via Object Tracking for Monocular 3D Object Detection: PLOT proposes a 3D pseudo-labeling framework that requires no auxiliary sensors or model retraining. By establishing cross-frame point-level correspondences via dense point tracking, decoupling camera ego-motion and object motion estimation from background/object trajectories, and introducing a global object memory to maintain long-term identity consistency, the framework aligns and fuses multi-view partial observations into complete object pseudo-point clouds. It achieves performance comparable to fully supervised methods on benchmarks like KITTI, KITTI-360, and Waymo, and generalizes well to non-driving scenarios.
PriorEye: Geospatial Visual Priors for End-to-End Autonomous Driving: PriorEye anchors offline-collected street-view images to planned routes to construct "geospatial visual priors." These priors are injected into end-to-end driving policies via a dual-memory architecture, equipping the agent with human-driver-like prospective perception while maintaining robustness under both sensor failures and prior corruption.
RESOLVE: A Multi-Resolution and Multi-Modal Dataset for Roadside Cooperative Perception: RESOLVE is the first real-world multi-resolution, multi-modal roadside cooperative perception dataset. By simultaneously deploying three LiDARs with 16/64/128 channels at real urban intersections while strictly maintaining sensor configuration and environmental consistency, it systematically evaluates the impact of LiDAR resolution variation on 3D object detection, multi-object tracking, and cooperative perception. It reveals how multi-modal fusion compensates for perception degradation under sparse point clouds and exposes the domain drift issue caused by the mismatch between training and inference resolutions.
SIMSplat: Language-Aligned 4D Gaussian Splatting for Driving Scenario Generation: SIMSplat embeds appearance, motion, and location language features directly into the nodes of a scene-graph 4D Gaussian representation, making a reconstructed driving scene queryable by free-form text and letting object-level editing, multi-agent path refinement, and VLM-driven scenario mining all run on that same representation; on Waymo it raises open-vocabulary object grounding accuracy from 0.35 to 0.73, reaches an 87.2% task completion rate with human guidance (71.1% fully VLM-driven), and cuts the multi-agent failure rate to 20.0%.
Sparse-Aware Vector Quantization for Bandwidth-Efficient Collaborative 3D Semantic Occupancy Prediction: VQSOP lets multiple vehicles exchange only codebook indices over V2X instead of dense 3D features: a confidence threshold first selects the voxels that actually carry information, a shared learnable codebook quantizes those voxel features to discrete indices by nearest-neighbor lookup, the receiver reconstructs them by table lookup and fuses them with the ego features, and a dual-branch (local convolution + dilated convolution) module adaptively refines the result for occupancy prediction — reaching 73.79% IoU / 41.54% mIoU on Semantic-OPV2V at 0.013 MB per frame.
SynFlow: Scaling Up LiDAR Scene Flow Estimation with Synthetic Data: SynFlow builds a motion-oriented synthetic data engine on CARLA and uses it to produce SynFlow-4k, 4,000 sequences (~940k frames) with noise-free per-point ground truth; a model trained on synthetic supervision alone transfers zero-shot to nuScenes/TruckScenes/Aeva, beats the fully supervised ∆Flow baseline (0.402) on TruckScenes by 31.8% at 0.274, and exceeds a from-scratch baseline trained on 4x the label budget after fine-tuning on just 5% of real labels.
2K Retrofit: Entropy-Guided Efficient Sparse Refinement for High-Resolution 3D Geometry Prediction: This paper proposes 2K Retrofit, a generic framework that enables existing 3D foundation models (e.g., Depth Anything, VGGT) to output 2K-resolution geometric predictions without any architectural modifications or retraining of the base models. The core idea is to utilize a frozen base model for fast coarse predictions, and then employ an entropy-guided sparse refinement mechanism to apply full-resolution corrections only to high-uncertainty pixel regions. This design achieves or exceeds the accuracy of full-resolution methods while significantly reducing computational and memory overheads.
ARC-Loc: Leveraging Azimuthal Ray Convergence as a Geometric Cue for Direct Cross-View Localization: Under a known heading or a bounded-noise heading prior, ARC-Loc converts ground-satellite matches into azimuthal lines and estimates position through differentiable weighted intersection and RANSAC, requiring neither depth estimation nor BEV transformation and reaching 2.52 m mean error on VIGOR Same-Area with known orientation, while remaining limited by narrow-view cross-area generalization.
BEVLM: Distilling Semantic Knowledge from LLMs into Bird's-Eye View Representations: BEVLM first establishes that fused BEV tokens support stronger spatial reasoning in language models, then trains the BEV encoder with frozen-LLM question-answering gradients and detection supervision, raising UniAD's NeuroNCAP score from 2.38 to 3.05 and reducing collision rate from 0.56 to 0.47 without introducing an LLM into driving-time inference.
CCFM: Collision-Constrained Flow Matching for Safety-Critical Scenario Generation: CCFM reformulates safety-critical driving scenario generation as flow matching sampling with hard collision constraints: it first dynamically selects feasible adversarial vehicles and collision types, and then progressively projects action sequences using contact point, relative heading, and severity constraints. It improves the collision rate to 46.4%/83.1% on nuScenes/nuPlan in an 80-frame closed-loop simulation, while achieving a collision type matching rate of 84.3%/85.6%.
CoGoal3D: Collaborative 3D Object Detection with 3D-Aware Fusion and Refinement: To tackle 3D spatial misalignment caused by disparate heights and attitudes among collaborators in V2X perception, CoGoal3D introduces multiscale 3D-Aware Global Fusion (3D-AGF) and Reconstruction-Guided Local Refinement (RGLR) coupled with Multi-Agent Collaborative Data Augmentation (MCDA), setting new SOTA performance across real-world datasets.
Comprehensive Robustness Analysis of LiDAR-based 3D Object Detection in Autonomous Driving: Addressing the critical flaw where standard evaluations over-rely on mAP while ignoring 3D physical and structural characteristics, this paper establishes a benchmark that decouples structural and predictive robustness, demonstrating that high-capacity voxel and transformer detectors suffer from systemic blindness under coordinated coordinate shifts.
Cooking beyond Frames: A Stereo Event Camera Dataset in the Kitchen: EventKitchen is the first large-scale, real-world, unscripted egocentric stereo event camera benchmark dataset capturing daily cooking activities across 13 diverse kitchens, featuring 5.5 hours of synchronized multimodal recordings, 10,762 action segments, 13,482 bounding boxes, and dense ground-truth depth maps across three core perception tasks.
CooperScene: Multi-Modal Cooperative Autonomy Benchmark with C-V2X Communication Characterization: CooperScene is the first collaborative autonomous driving dataset that synchronously captures C-V2X communication characteristics (throughput, latency, packet loss rate, jitter) and multi-modal sensor data (LiDAR, camera, GNSS/IMU) in real-world road scenarios. It covers interactions among three intelligent vehicles and one road-side unit (RSU) across diverse scenes such as intersections, highway ramps, and parking lots. Benchmark evaluations on this dataset reveal that the mAP of existing cooperative perception methods drops by up to 49% under realistic C-V2X bandwidth constraints compared to ideal network assumptions, highlighting the massive gap between idealized evaluations and deployable performance.
Counting Trees from Satellite Imagery with Noisy Supervision: Addressing ambiguous crown boundaries in dense canopies and noisy airborne LiDAR annotations, TreeMatch models satellite tree counting as Unbalanced Optimal Transport (UOT) density matching and bootstraps transport marginal residuals to iteratively self-correct noisy labels, outperforming detection and regression baselines on the multi-continent TinyTrees benchmark.
Cross-View Yaw Estimation in Location Uncertainty with Line-Aligning Yaw Scoring: Addressing the failure of cross-view localization under large yaw ambiguity and location noise, LAYS decouples yaw from translation by matching vertically aggregated ground columns with BEV pixels, leveraging proven radial invariance in a 3D voting space to achieve sub-degree yaw accuracy without ground height assumptions.
DecoupleGS: Interactive 3D Gaussian Splatting for End-to-End Autonomous Driving Testing: Addressing the core trilemma among fidelity, interactivity, and real-time throughput in closed-loop end-to-end autonomous driving simulation, DecoupleGS explicitly decouples driving scenes into static backgrounds and canonical dynamic assets, resolving efficiency, geometric, and photometric conflicts via asset compression, map-guided registration, and proxy-based relighting.
DeGuNet: Depth-Guided Ultra-Compact Backbones for Efficient LiDAR-Camera 3D Detection: DeGuNet replaces parameter-heavy, 2D-pretrained visual backbones with an ultra-compact 0.31M depth-guided backbone that leverages masked partial convolutions (MPIR) and masked attention (MMViT) to prevent sparse LiDAR projection contamination, reducing GPU memory by 66.5% and accelerating inference while gaining up to 6.20 mAP across established multi-modal 3D detectors.
Delaunay Canopy: Building Wireframe Reconstruction from Airborne LiDAR Point Clouds via Delaunay Graph: Delaunay Canopy constructs a Delaunay graph over airborne LiDAR point clouds to compute face normals, edge dihedral angles, and vertex corner scores as explicit curvature signatures, establishing an adaptive 3D search space for precise corner and wire selection while overcoming the search explosion of 3D baselines and the loss of internal corners in 2D heightmap projections.
DH-VLM: Dual-Horizon Cooperative Latent Reasoning for Autonomous Driving: Proposes DH-VLM, an asymmetric vehicle-infrastructure cooperative latent reasoning framework where an infrastructure large model aggregates multi-layer hidden states into global cognitive guidance, conditionally refining on-board ego planning via an infrastructure-driven latent evolution mechanism to cut communication costs by 57.3% while significantly enhancing long-horizon safety and planning accuracy.
DriveWeaver: Point-Conditioned Video Inpainting for Controllable Vehicle Insertion in Autonomous Driving Simulation: DriveWeaver proposes a video diffusion inpainting framework using vehicle point cloud renderings as pixel-level conditions. By employing a lightweight PointAdapter to inject geometric conditions into a frozen Wan2.1 diffusion transformer backbone, it generates foreground vehicles with consistent lighting integration and precise geometry. Additionally, a global-local hierarchical inpainting strategy is designed to eliminate long-horizon drift, and the generation results can be further distilled into 3D Gaussian Splatting for real-time rendering.
ECoSim: Data Efficient Fine-Tuning for Controllable Traffic Simulation: ECoSim freezes existing traffic generators and adds sketch, behavior-latent, or text control through identity-initialized FiLM adapters and closed-loop training, reducing SMART sketch-control mADE from 1.5044 to 0.2561 with only 1% paired control data while using context matching to transfer behaviors into new traffic scenes.
ECTraj: Enhanced Consistency Training for Multi-Agent Trajectory Prediction: ECTraj introduces an enhanced conditional Consistency Model pipeline for multi-agent trajectory prediction that couples teacher waypoint ground-truth fusion with training-time top-K consistency matching, achieving single-step generation with a 10x inference speedup and superior accuracy over diffusion baselines.
Explainability-aware Frustum Attack: Exposing Structural Vulnerabilities in LiDAR-Based 3D Object Detectors: This paper introduces SALL, an explainability framework that aggregates instance-level Integrated Gradients across driving scenes to derive universal class-level saliency maps for LiDAR 3D detectors, and proposes EFA, a geometry-consistent frustum attack that selectively perturbs only critical frustums to drop detection recall to near zero while reducing required perturbation budgets by 25%–50%.
FDR-Occ: Factorized Dense Routing for Full-Spectrum 3D Occupancy Prediction: Addressing the severe locality bottleneck and rigid calibration dependence of conventional ray-based view transformation, FDR-Occ introduces Factorized Dense Routing (FDR) and a Resolution-Context Decoupled Architecture, achieving sub-quadratic global reachability while maintaining extreme robustness even when physical camera extrinsics are withheld.
FLM-Occ: Feed-forward Likelihood Maximization for Efficient Indoor Occupancy Prediction: FLM-Occ redefines indoor monocular occupancy prediction from voxel classification to voxel distribution estimation. By training the network with Maximum Likelihood Estimation (MLE) to predict the mixture model parameters of the scene in a feed-forward manner, it outperforms the previous state-of-the-art method (which requires 1200 Gaussian primitives) with only 32 superquadric primitives, while achieving a \(3.7\times\) speedup in inference.
FreeGen: Feed-Forward Reconstruction–Generation Co-Training for Free-Viewpoint Driving Scene Synthesis: FreeGen establishes a feed-forward reconstruction-generation co-training framework unifying 3D Gaussian Splatting and video diffusion, delivering consistent and realistic free-viewpoint driving scene synthesis from single-trajectory image inputs without per-scene optimization.
Geometry-Aware Spatio-Temporal Context Modeling for 4D Occupancy Forecasting: GAST establishes an end-to-end continuous BEV paradigm for 4D occupancy forecasting that replaces discrete autoregressive tokenization with pose-driven progressive explicit-implicit generation and dual-path spatio-temporal context modeling, outperforming prior SOTA on Occ3D-nuScenes by 7.67% mIoU with a 2.84x speedup.
Geometry-Preserving in 3D Gaussian Splatting for LiDAR-Camera Extrinsic Calibration: This work proposes GeoP-Calib, which mitigates the "geometric decay" issue in 3DGS calibration through two designs: Dense Depth Anchoring (DDA, which aggregates multi-view LiDAR point clouds to establish a dense depth prior) and Gradient Decoupling (GD, which blocks the backpropagation of photometric loss to Gaussian spatial parameters). It achieves translation errors significantly superior to existing targetless calibration methods on KITTI-360 and KITTI.
HilDA: Hierarchical Distillation with Diffusion for Advancing Self-Supervised LiDAR Pre-training: HilDA distills LiDAR backbones using the "hierarchical semantic structure + CLS global context" from Vision Foundation Models (VFMs) rather than just the final layer features. It additionally incorporates a conditional diffusion task predicting future BEV occupancy to complement spatio-temporal geometric information, setting a new SOTA across multiple downstream tasks in cross-modal distillation: segmentation, detection, scene flow, and semantic occupancy.
Honey, I Shrunk the Arc de Triomphe!: To eliminate the severe "scale-collapse" flaw where monocular geometry foundation models metrically underestimate distant structures and landmarks in the wild, this work builds MetricScenes—the first large-scale real-world metric dataset anchored by geo-tagged street-view imagery and physical stereo baselines—and develops a two-stage edge-aware Poisson depth completion strategy, yielding WildMoGe which restores faithful absolute physical dimensions without sacrificing standard benchmark accuracy.
Horizon3D: Sparse Radar-Camera Fusion for Long-Range 3D Perception in Autonomous Driving: Horizon3D proposes a sparse radar-camera fusion framework for long-range autonomous driving perception. It utilizes Gaussian primitives to capture object-level details and sparse BEV features to encode scene-level context. Additionally, it addresses the dilemma of long-range sparsity and high-speed large displacements simultaneously through dual-path temporal fusion (BEV path for multi-frame accumulation + Gaussian path for object-by-object motion propagation). On TruckScenes, it outperforms the previous state-of-the-art radar-camera fusion method by +3.0 NDS / +1.6 mAP, while achieving faster inference speed than existing BEV fusion methods.
HSDF-Lane: Height-Aligned Signed Distance Field with Semantic Lane Prior for 3D Lane Detection: HSDF-Lane replaces traditional sparse slope anchors with a vertically-aligned signed distance field (HSDF) to implicitly model road surfaces. Through differentiable rendering, it simultaneously outputs accurate height maps and surface-aligned features, while introducing Lane-aware Semantic Positional Encoding (LSPE) to inject lane existence priors into Transformer queries. This approach achieves dual SOTA in both 3D lane detection and height estimation on the OpenLane benchmark.
Learning 1-Bit LiDAR-based Localization with Auxiliary Objective: BiLoc is the first 1-bit binary neural network framework designed for 6-DoF outdoor LiDAR localization; by optimizing an information-bottleneck auxiliary objective with soft-masked channel distillation and pose-aware optimal transport during training only, it cuts Oxford average position error to 7.56 m (a 10.11% reduction over prior 1-bit SOTA) and achieves a 2.07x latency speedup on an RTX 5090 without any inference overhead.
Learning to Suppress SPAD-based LiDAR Flare: Addressing the severe point cloud distortions and spurious obstacles caused by pile-up and optical crosstalk in SPAD LiDARs facing retro-reflective surfaces, this paper reformulates flare suppression as a semantic segmentation task, introduces a physically-informed framework (PILF) that integrates dual echoes with ambient illumination, and restores undistorted scene geometry via localized echo replacement.
mmIR: Frequency-Space Inverse Rendering for 3D Millimeter-Wave Radar ADC Synthesis: Addressing the severe scarcity of high-resolution 3D mmWave radar data and the ill-posed nature of raw ADC inversion, mmIR presents the first open-source, end-to-end differentiable FMCW radar inverse renderer that leverages LiDAR-derived meshes as geometric scaffolds to jointly optimize per-vertex ITU physics materials, vertex normals, and antenna beam patterns, enabling high-fidelity cross-sensor transfer and high-resolution 3D RAE occupancy synthesis via dense virtual apertures.
ModTrack: Sensor-Agnostic Multi-View Tracking via Identity-Informed PHD Filtering with Covariance Propagation: ModTrack restricts neural networks strictly to per-sensor detection and feature extraction, executing all downstream fusion, association, and temporal filtering through closed-form analytical methods—including Jacobian covariance back-projection, \(\chi^2\) graph clustering, precision-weighted fusion, and an identity-informed GM-PHD filter—achieving 95.5 IDF1 and 91.4 MOTA on WildTrack with seamless cross-modal deployment.
MOJITO: Modal Joint Learning for Unified End-to-End Autonomous Driving: MOJITO presents a fully unified sensor-to-action end-to-end autonomous driving framework that eliminates the information bottleneck of cascaded architectures via block-wise Modal Joint Attention, achieving state-of-the-art continuous trajectory diffusion generation without predefined trajectory anchors or dense auxiliary supervision.
MV2GF: Multi-view Pedestrian Detection with a Visual Geometric Foundation Model: Addressing the failure of visual geometry reasoning and severe overfitting to perspective distortion patterns under unseen camera configurations in conventional multi-view pedestrian detection, MV2GF introduces Depth Anything 3 to fuse task-specific pedestrian semantics with foundational geometric representations and aggregates features into 3D voxels using metric 3D pointmaps, achieving state-of-the-art cross-camera generalization.
NegROI: Click-Centric Uncertainty-Guided Refinement with Scene-Conditioned Negative Prompts for Robust Interactive 3D Segmentation: NegROI addresses boundary under-segmentation and persistent hard false positives in interactive point cloud segmentation by introducing click-centric multi-resolution ROI refinement alongside scene-conditioned negative prototype learning, significantly boosting click efficiency and outdoor LiDAR transfer robustness.
OccDirector: Language-Guided Behavior and Interaction Generation in 4D Occupancy Space: OccDirector introduces the first purely natural-language-guided 4D occupancy behavior and interaction generation framework, removing dependencies on rigid geometric trajectories by leveraging a frozen VLM encoder, a dual-stream spatio-temporal MMDiT, and a history-prefix anchoring strategy to synthesize physically plausible, causally consistent multi-agent dynamic rollouts.
Open-Vocabulary BEV Segmentation with 3D-Aware Geometric Constraints: This work proposes the first open-vocabulary BEV segmentation framework, OVBEVSeg. By reversing the geometric reasoning path (establishing 2D-BEV correspondence with unsupervised 3D detection \(\rightarrow\) optimizing geometry via BEV-aware 3DGS \(\rightarrow\) distilling to a feed-forward network), it lifts VLM semantics from 2D to the BEV space, outperforming closed-set methods by 15.3 mIoU on novel classes in nuScenes with faster inference speed.
Open-Weather Robust 3D Detection via Dual-Critic Diffusion Alignment: Addressing LiDAR degradation under adverse weather and the closed-world limitations of existing fusion models, this paper introduces Dual-Critic Guided Diffusion Alignment (DCDA), which leverages 4D radar features to progressively restore degraded LiDAR representations toward a clean manifold without requiring paired clean-corrupted data or explicit weather labels, significantly boosting 3D detection generalization across unseen weather types and severities.
PersonaDrive: Controllable Trajectory Prediction with Multi-Dimensional Driving Personas: PersonaDrive establishes a multi-dimensional driving persona framework by decomposing driving behavior into a 3×3 grid of Temporal Urgency and Ride Comfort, enabling natural-language controllable trajectory prediction via hierarchical anchor transformation (PCAT), multimodal gated fusion (PCMF), and axis-aligned ordering and diversity objectives.
Pixel-wise Geo-registration of Drone and Satellite Images: Addressing the fundamental limitations of prior cross-view localization confined to image-level camera center estimation and planar homographies failing under severe parallax, SkyReg introduces a feed-forward 3D geometry-aware registration framework alongside the first pixel-wise geodetic benchmark SkyReg-Bench, establishing robust, high-accuracy per-pixel GPS coordinate prediction.
Plug-and-Play Traffic Element Awareness for End-to-End Autonomous Driving: Addressing the widespread neglect of regulatory traffic lights and road signs in mainstream end-to-end autonomous driving systems, this paper introduces a lightweight, plug-and-play auxiliary 3D traffic element supervision along with a language-guided ego-centric topology conditioning mechanism, consistently boosting safety and compliance across nuScenes, NAVSIM-v1/v2, and Bench2Drive.
RAF: Reliability-Aware Fusion of Camera, LiDAR, and 4D RADAR for Robust 3D Object Detection in Adverse Weather: Addressing sensor degradation where adverse weather thins out point clouds while unevenly occluding camera lenses, RAF uses weakly-supervised cross-modal feature alignment and calibration-aware local matching to produce pseudo reliability labels that explicitly gate camera features, achieving up to +6.5 APBEV and +7.4 AP3D gains over LiDAR–4D RADAR baselines on K-Radar.
Real-Time Source-Free Object Detection: This paper constructs RT-SFOD, a real-time source-free domain adaptive object detection framework based on YOLOv10 (the first NMS-free dual-head detector). It proposes Dual-Head Pseudo-Label Fusion (DHF) to recover objects missed by the O2O head while maintaining high precision, alongside a Multi-Scale Adaptive Representation Diversification Loss (MARD) to combat feature rank degradation caused by domain shift. On four domain shift benchmarks, it achieves a 1.4-3.5% mAP improvement with 1.3x faster inference speed and ~2x parameter compression.
RECO: Region-Aware Compensation for Extrinsic Perturbations in Roadside 3D Detection: To tackle severe feature misalignment caused by roadside camera vibrations and structural deformations, RECO predicts a learnable near/far range boundary alongside piecewise 6-DoF pose offsets to compensate for nominal extrinsics, smoothly blending geometries via a differentiable sigmoid gate and auxiliary reprojection loss to achieve superior 3D detection robustness under both transient and persistent extrinsic perturbations.
ReinDriveGen: Reinforcement Post-Training for Out-of-Distribution Driving Scene Generation: ReinDriveGen couples dynamic point cloud completion with a condition-guided video diffusion model and introduces an RL-based post-training paradigm with a dedicated pairwise preference model and pairwise win-rate rewards, substantially enhancing the photorealism of out-of-distribution safety-critical driving scenarios without ground-truth supervision.
RePL: Pseudo-label Refinement for Semi-supervised LiDAR Semantic Segmentation: Addressing error propagation and confirmation bias from noisy pseudo-labels in outdoor semi-supervised LiDAR semantic segmentation, RePL detects unreliable voxels via teacher-student agreement and actively reconstructs them with a learnable mask token, setting new state-of-the-art results on nuScenes-lidarseg and SemanticKITTI without per-benchmark hyper-parameter tuning.
Representations Before Pixels: Semantics-Guided Hierarchical Video Prediction: Re2Pix decouples future video prediction into two stages—vision foundation model (VFM) semantic representation forecasting followed by semantics-guided latent diffusion synthesis—and introduces nested dropout alongside mixed supervision to resolve autoregressive train-test mismatch, delivering 7× to 14× training convergence speedups and state-of-the-art temporal semantic consistency and visual fidelity on driving benchmarks.
ScenarioControl: Vision-language Controllable Vectorized Latent Scenario Generation: ScenarioControl introduces the first vision-language controllable vectorized latent diffusion framework for 3D driving scenario synthesis, combining a cross-global control mechanism with collision regularization and cardinality regression to generate realistic 3D road layouts, reactive traffic agents, and sensor-level video continuations from text prompts or single forward-facing dashcam images.
SEM-ROVER: Semantic Voxel-Guided Diffusion for Large-Scale Driving Scene Generation: SEM-ROVER introduces \(\Sigma\)-Voxfield, a discrete surface representation with fixed-cardinality surface samples, and scales local 1D transformer diffusion to 100,000 \(m^2\) driving scenes via progressive outpainting and feed-forward deferred rendering within 8 GB VRAM.
Sen-Cap: Sensor-Flexible and Noise-Resilient Human Motion Capture via LiDAR-Camera Integration: Sen-Cap presents a calibration-free 3D human motion capture framework that supports plug-and-play arbitrary numbers of LiDARs and cameras while remaining resilient against heavy point cloud clutter, estimating local poses via human-centric feature alignment with bottleneck attention and tracking global translation via iterative offset shrinkage.
SIGMA-Lane: Scale-pyramId Gated MAmba for Temporally Consistent Video Lane Detection: To tackle state contamination and temporal flickering caused by heavy vehicle occlusions in streaming video lane detection, SIGMA-Lane introduces an SSM-consistent dual-gating mechanism that gates both the state-write and residual-fusion paths, coupled with coordinate-consistent affine alignment and structural spatial retrieval to achieve superior temporal consistency and efficiency.
SP-TransientBench: A Real-Captured Single Photon Perception Benchmark: The first real-captured multi-task single-photon LiDAR (SPL) 3D perception benchmark, providing full \(256 \times 192 \times 672\) transient time-of-flight histograms, calibrated poses, and histogram-domain multi-return semantic annotations across depth estimation, multi-view reconstruction, and 3D semantic segmentation.
SWAN: World-Aware Adaptive Multimodal Networks for Runtime Variations: Addressing concurrent runtime variations in sensor quality of information, platform compute constraints, and sample complexity in autonomous driving, SWAN unifies a differentiable NeuralSort-based QoI allocation controller, an input-conditioned SkipGate module, and DETR token pruning, reducing compute by up to 49% with minimal accuracy degradation.
Targeted Structure Completion for Sparse-View 3D Reconstruction in Autonomous Driving: Addressing the substantial computational redundancy of uniform volumetric densification in sparse surround-view driving scenarios (<15% overlap), FocusGS decouples deterministic surfaces from ambiguous boundaries via a 3D geometric ambiguity manifold, instantiating and optimizing queries exclusively within this sparse topological subspace to cut Gaussian count by ~74% and rendering time by ~34% while establishing state-of-the-art visual quality.
TASE: Truncation-Aware Semantic Embeddings for 3D Scene Understanding and Editing: TASE projects pretrained 2D visual features into a truncation-aware, positional-bias-free latent space via a Matryoshka autoencoder and lifts them into 3DGS, enabling text-driven 3D scene editing with explicit control over semantic abstraction and substantial geometric changes.
Towards Metric-Agnostic Trajectory Forecasting: This paper advocates that trajectory forecasting should only learn a "well-calibrated predictive distribution," treating conflicting metrics like minFDE and soft mAP as downstream tasks. By utilizing a set of retraining-free sampling policies (TraDiE policies) during evaluation, the optimal \(K\) trajectories for each metric are extracted from the same distribution, allowing a single DONUT-NLL model to simultaneously achieve SOTA on both distance-based and window-based metrics on Waymo.
Towards Practical Lossless Neural Compression for LiDAR Point Clouds: To tackle high-resolution contextual sparsity and entropy coding collapse caused by cross-platform floating-point discrepancies in high-precision LiDAR point cloud compression, this paper introduces a lightweight octree predictive coding framework integrating Geometry Re-Densification (GRED) and Cross-Scale Feature Propagation (XFP), paired with an end-to-end integer-only inference pipeline that achieves bit-exact cross-platform decoding and 14 FPS real-time compression.
Towards Robust Driving Perception: A Flexible Scale-Driven Family for Self-Supervised Monocular Depth Estimation: To tackle photometric reprojection corruption caused by moving objects and boundary blurring in lightweight models within autonomous driving, FlexDepth establishes a flexible model family scaling from Nano to X-Large. It introduces a pre-convolution-free Scale-Driven Decoder alongside a two-stage static-dynamic decoupled training strategy derived from cross-epoch prediction variance, securing SOTA accuracy and ultra-low edge latency without requiring semantic or optical flow priors.
Trajectory-aware Cross-view Geo-Localization with Sequential Observations: To tackle the lack of linguistic route narratives and allocentric spatial layout awareness in cross-view geo-localization, this paper introduces SeqGeo-VL comprising ~39K video-text-satellite triplets and proposes TrajLoc, a unified framework leveraging two-stage curriculum learning and a lightweight TrajMod module that significantly enhances both video and text retrieval.
TraversRL: Traversable Pedestrian Pathway Generation With Reinforcement Learning: Addressing the fragmentation problem of traditional aerial image segmentation for sidewalk mapping, TraversRL models pedestrian network generation as an iterative sequential decision process from a traveler's perspective, employing a discretized direction-distance action space optimized with GRPO using global graph-level IoU and local stepwise geometric efficiency reward shaping, which more than doubles the connectivity metric TravSim while significantly pruning redundant branches.
UECP: Uncertainty-Enhanced Collaborative Perception: Addressing the issue where detector-coupled confidence maps trigger false-positive amplification and feature corruption in collaborative perception, UECP establishes physically grounded uncertainty maps from raw LiDAR point densities and aggregates multi-agent BEV features via uncertainty-weighted downsampling and residual fusion.
Understanding Cross-Rig Generalization in Automotive Perception: a Multi-Rig Benchmark and Rig Variation Metrics: This paper constructs PCCR, a simulation benchmark containing 14 systematically designed camera rigs, and proposes two geometric descriptors based on calibration parameters: Rig Variance (RigV) and Rig Contrastive Distance (RigCD). Experiments demonstrate that camera translation discrepancy is the dominant factor in cross-rig generalization performance degradation, and RigCD can predict the ranking of transfer difficulty between different rigs, achieving a Spearman rank correlation of up to 0.80.
UniTeD: Unified Temporal Diffusion for Joint Perception and Planning in Autonomous Driving: UniTeD integrates the anchor queries of perception (detection, mapping, motion prediction) and planning into a unified diffusion denoising process for joint modeling. It utilizes a Temporal Transition Module (TTM) to align the noise levels of historical and current frames, and employs an Anchor Refresh Strategy (ARS) to alleviate the training-inference distribution shift in sparse diffusion. Using only cameras, it outperforms discriminative and planning-only diffusion methods on NAVSIM and Bench2Drive, achieving 90.2 PDMS / 87.3 DS.
Unordered Landmark Visual Navigation: Addressing the severe perceptual aliasing and mapping failures caused by the lack of temporal and odometric priors in unstructured image sets, ULVN establishes the first unified RGB-only unordered landmark visual navigation framework, combining calibrated geometric verification with maximum spanning forest refinement for mapping, 2D graph belief propagation for localization, and bottleneck-aware path search for closed-loop execution.
Unsupervised Point Cloud Registration via Training-Time Semantic Guidance: Addressing the issue where geometric ambiguity in outdoor scenes causes noisy pseudo-labels and severe "semantic collapse" during unsupervised point cloud registration, CAESAR leverages a frozen off-the-shelf 3D segmentation network solely during training as an external semantic anchor—combined with dual-cue candidate reselection, lightweight batch-specific teacher refinement, and semantic predictive distillation—achieving state-of-the-art registration performance on KITTI and nuScenes with zero test-time overhead.
Unveiling Transferability in Trajectory Prediction via Latent Scene Embeddings: This paper proposes the TransferAtlas framework, which employs a Graph Neural Network encoder to map 24 trajectory prediction datasets jointly into a 32-dimensional unified latent space. In this space, each dataset is represented as a Gaussian distribution. The KL divergence between two distributions quantifies the difficulty of transferring from the source domain to the target domain. This achieves a Spearman rank correlation coefficient of 0.811 for zero-shot transfer, offering principled guidance for dataset selection, pre-training source recommendation, and foundation model construction.
VIPS: Vehicle-Infrastructure Cooperative Planning Benchmark via Pseudo-Simulation: Introduces VIPS, the first pseudo-simulation benchmark for vehicle-infrastructure cooperative planning evaluating robustness and error recovery via synchronized real and perturbed novel-view observations, paired with CoS-V2X, an efficient sparse planning framework.
When the City Teaches the Car: Label-Free 3D Perception from Infrastructure: This paper introduces an infrastructure-taught, label-free 3D perception paradigm where stationary roadside units (RSUs) learn local 3D detectors from unlabeled LiDAR data via background persistence and broadcast predictions to passing ego vehicles as pseudo-labels, training an ego detector that operates completely standalone at test time with 82.3% vehicle AP on the CARLA-based CIVET benchmark.
Adaptive Neural Dynamics for Robust Geometric LiDAR-Inertial State Estimation on UAVs: NeuroSDE-LIO retains the geometric estimation backbone of FAST-LIO2, adaptively combines linear bias priors with neural stochastic bias dynamics, and propagates their uncertainty into the filter, reducing UAV-Lab absolute pose error from 2.13 m to 1.34 m at a LiDAR processing rate of 19.2 Hz.
Agent-OBJ: Prompt-Driven 3D Adversaries for Multi-Modal Perception: Agent-OBJ searches language prompts controlling the shape and appearance of a 3D object using only detection boxes and confidence scores, combining naturalness and cross-view stability constraints to achieve 64.0% joint disappearance from camera and fusion detectors in simulated nuScenes insertion experiments.
AiSCREAM: Absolute Target Localization with Language-Conditioned Cross-View Alignment for Autonomous Vehicles: AiSCREAM uses an aerial image generated from a single front view as an auxiliary spatial cue, aligns the two views through language-conditioned semantic segmentation, and combines multitask localization with MLLM refinement to predict a metric target and acceptable region, reducing DRAMATiST RMSE from the strongest model baseline's 8.56 m to 5.00 m.
Analytic Bayesian Uncertainty for LiDAR Segmentation: A Single-pass Generative Approach: GMM-NIG places conjugate Bayesian posteriors over the means and variances of a class-conditional Gaussian mixture head attached to a deterministic segmentation backbone, then uses analytic Student-t prediction to reach 65.71% mIoU, 2.01% ACE, and 53.33% failure-detection AUPR with RangeFormer on SemanticKITTI without repeated stochastic forward passes.
ASTAD: Asymmetric Style Transfer for Auto-driving Synthetic-to-Real Domain Adaptation: Addressing the asymmetric constraint in autonomous driving scenarios where "synthetic images have perfect annotations, while real style reference images have no annotations," ASTModel proposes a training-free two-stage framework. It first extracts a coarse semantic prior from unlabeled style images using DINO prototype matching. Then, during the reverse diffusion process, it refines the prior via multi-layer semantic voting, uses robust median threshold filtering and pixel-proportion modulated hybrid AdaIN to achieve class-consistent style injection, generating realistic target-domain style data with a 3.2x speedup while preserving semantic structures.
AV2T-Gen: Aerial Visible to Thermal Generation with Environment and Vehicle State Guidance: AV2T-Gen adds environmental prompts and motion-aware object conditions to Instruct-Pix2Pix-style diffusion so that aerial RGB-to-thermal translation goes beyond appearance transfer, reducing ThermalGen's FID from 83.16 to 67.45 in joint four-scene evaluation while improving vehicle detection on synthetic thermal images.
BEVOpen3D: Towards Open-World 3D Object Detection in Bird's-Eye-View: BEVOpen3D teaches a vision model to correct rule-generated 3D boxes using seen-class supervision, then trains a LiDAR-only student through triple-source pseudo-label fusion and BEV heatmap distillation, reaching 30.10 mAP and 16.26 unseen AP on the nuScenes seven-novel-class split, gains of 0.73 and 1.33 percentage points over Find n' Propagate.
BitRIC: Efficient Neural Compression of LiDAR Range Images via Hierarchical Bitplanes: BitRIC decomposes integer range images into binary planes ordered from most to least significant, entropy-codes them using shared context and three-stage probability estimation, and compensates for truncated low bits through one-step residual refinement, achieving D1 BD-BR of −61.73% against JPEG2000 and lossless coding at 6.05 Bpp on WOD while retaining strong downstream detection accuracy.
Boba: Batched Simulation for Physics-Based Gaussian Digital Twins: Boba jointly optimizes the physical model, skinning, and GPU execution of PhysTwin-based deformable Gaussian digital twins, reducing single-instance Orin latency from 330.9 ms to 32.5 ms and reaching 3309.9 aggregate FPS on RTX 4090 while retaining reconstruction and prediction quality close to the baseline.
CaRe: Critical Parameter Rectification for Efficient Visual Modeling: CaRe expresses a standard weight as a dense base plus a learned Softplus-gated correction, concentrating training-time corrections on important entries and merging them into an ordinary layer before deployment; CaReNet-M1 achieves 73.6% ImageNet-1K Top-1 at 33,203 images/s on an A100, outperforming EfficientViM-M1 by 0.7 percentage points.
CGCC: Towards Generalizable Clothes-Changing Person Re-Identification: To tackle poor generalization in real-world clothes-changing person re-identification, this paper constructs CGCC—a large-scale synthetic dataset (4,101 IDs, 217k images) featuring global cultural/spatiotemporal diversity and structured semantic annotations—and proposes the TGIR framework that projects visual features orthogonally away from an SVD-modeled identity-irrelevant semantic subspace.
DiverseAD: A Large-Scale Driving Dataset with Diverse Atmospheric Conditions: DiverseAD provides both a 150K-scene driving dataset covering diverse atmospheric conditions and a dual-anchor feature-learning method: trajectory direction patterns and relationships from frozen DINOv3 features guide visual representations, reducing average L2 trajectory error on DiverseAD from 1.44 m for fine-tuned FSDrive to 1.24 m.
Doe-2: 3D Representation World Model for Unified Driving Scene Forecasting: Tsinghua University and Yinwang Intelligent propose Doe-2, a native 3D driving world model that couples action-conditioned autoregressive spatio-temporal dynamics with a lightweight 3-layer DiT refinement head in a compact 3D latent space, enabling single-pass decoding of multi-view RGB, metric depth, 2D semantics, and 4D occupancy grids from camera-only inputs.
DRS-VPT: Directly Re-localizing in Scenes using a Vision and Point Transformer: DRS-VPT presents the first foundational feed-forward transformer architecture for cross-modal image-to-scan registration, fusing unaligned RGB query images and raw uncolored 3D point clouds via alternating global-local attention and unrolled differentiable direct reprojection alignment to achieve robust, initialization-free 6-DoF localization across diverse driving and indoor environments.
EgoTraj: Real-World Egocentric Human Trajectory: EgoTraj is the first open multimodal egocentric human trajectory dataset and benchmark collected with consumer AR headsets (Meta Quest Pro) in real-world outdoor urban environments, capturing synchronized 6DoF head poses, 3D gaze vectors, RGB videos, and VLM scene descriptions across 75 subjects navigating self-chosen routes, where cascaded cross-attention demonstrates the decisive predictive gain of gaze and scene priors.
ESNE: Efficient Surface Normal Estimation for LiDAR Point Clouds with Sequential Modeling and Variability Guidance: For large-scale sparse, non-uniform LiDAR point clouds, ESNE develops an ultra-efficient architecture combining state space models with 1D sequential local attention over Hilbert curves and attention-derived structural variability feature refinement, reducing model parameters by 89.4%, doubling inference speed, and reducing mean angular error by 33.4%.
Event-LiDAR: 3D Eventification for Efficient Point Cloud Processing: Addressing the severe spatial-temporal redundancy in multi-scan LiDAR perception, Event-LiDAR introduces a training-free 3D eventification front-end that approximates short-term geometric evolution using local first-order geometric models and extracts unpredictable signed geometric deviations as 3D events, cutting over 70% of points while preserving full-scan detection accuracy with up to a 23% end-to-end inference speedup.
Fast and Scalable LiDAR Data Generation for Autonomous Driving Simulation without Raycasting: GnoSphere360 completely replaces computationally expensive ray-mesh intersections by projecting triangular meshes onto a sensor-centered unit sphere, binning them onto an azimuth-elevation grid, and resolving exact depths via gnomonic projection, achieving a 35x speedup over native CARLA and outpacing NVIDIA OptiX by up to 3.65x in ultra-dense LiDAR scenarios.
FrozenDrive: Zero-Shot Text-Guided Driving Scene Generation and Data Augmentation: FrozenDrive completely freezes the pretrained Stable Diffusion backbone with zero added parameters, achieving both multi-view and temporal consistency purely by rearranging the inputs of self-attention (concatenating multi-view features and injecting reference frames). This preserves the original diffusion prior to enable zero-shot text-guided driving scene generation. Directed synthesis of rare adverse weather data (e.g., night, rain, snow) via text prompts significantly enhances downstream perception and planning robustness.
HVA-Fusion:Hierarchical Velocity-Aware 4D Radar-LiDAR Fusion for Robust 3D Object Detection: Addressing the severe degradation of LiDAR in adverse weather and the sparsity and non-representative Doppler radial velocity of 4D radar, this paper proposes HVA-Fusion, a hierarchical velocity-aware fusion framework that estimates true motion velocity, performs pillar-level adaptive interaction, densifies features via RCS-guided Gaussian diffusion, and fuses BEV features using multi-scale bidirectional deformable attention gates.
Hybrid Event-Frame Sensors: Unified Noise Modeling, Calibration, and Simulation: Proposes the first unified statistical noise model and calibration pipeline for hybrid event-frame sensors, and constructs H-ESIM, a simulator capable of generating RAW frames and event streams with noise distributions consistent with real sensors. Its sim-to-real transferability is validated on video frame interpolation and deblurring tasks.
Ice Cloud Geometry Retrieval with Calibrated Uncertainty from Passive Satellite Imagery: Addressing the coverage gap between narrow active radar/lidar nadir tracks (~1.6% supervision) and wide passive satellite swaths lacking direct vertical sensitivity, this paper introduces an ERA5-conditioned ConvNeXt-UNet that produces dense predictions of eight ice cloud geometry and microphysics targets with guaranteed 90% coverage via conformalized quantile regression (CQR), demonstrating that the retrieval operates as an independent per-pixel spectral mapping rather than spatial interpolation.
Label-Free Text Prototype Adaptation for Open Vocabulary Segmentation: VPTA introduces a label-free text prototype adaptation framework for open-vocabulary segmentation that keeps all model weights frozen, using only unlabeled target-domain images to rectify the language prior bias of CLIP text prototypes via margin-based pixel filtering, class reliability-scaled updates, and reliability-descending orthogonalization.
LEO-Fuse: A Modality- and Task-Agnostic Universal Framework for Multimodal Human Sensing: Addressing sensor dropouts and modality collapse in non-intrusive human sensing, LEO-Fuse introduces a universal Local Expert Orchestration framework with frozen unimodal encoders, mask-aware transformer fusion, and local adaptive routing to handle all 31 non-empty subsets across five modalities with a single model.
Noise is a Good Teacher: A Noise-Driven Framework for Robust Collaborative Perception: Addressing the coupled challenges of pose misalignment and bird's-eye-view (BEV) feature corruption in collaborative perception, this paper proposes NoiseGT, a noise-driven framework that leverages artificial noise as a self-supervised teacher via multi-scale high-noise feature denoising and multi-level pose disturbance injection, achieving state-of-the-art 3D detection robustness without requiring clean intermediate supervision.
OmniNWM: Omniscient Driving Navigation World Models: OmniNWM bridges the "state-action-reward" triad of driving world models within a unified probabilistic framework: it jointly generates pixel-aligned panoramic RGB / semantic / depth / 3D occupancy, decouples ego-motion from camera assembly geometry using normalized panoramic Plücker ray graphs to achieve zero-shot cross-camera transfer, and directly derives dense rewards from the generated occupancy volume, enabling closed-loop evaluation of planning agents in the generated world.
PointLAM: Local Attentive Mamba for Efficient Point-based 3D Object Detection: Addressing the longstanding dilemma between high computational latency in point-based architectures and quantization loss in voxel-based detectors, PointLAM introduces the Laplacian Point Sampler (LPS) with doubly sorted sampling and the Local Hadamard Aggregator (LHA) with transient grid routing coupled with bidirectional Mamba, achieving top-tier accuracy and voxel-rivaling inference speed while strictly preserving fine-grained point geometry.
PriorMaskMap: Robust Online Vectorized Map Construction with Biased Priors: Addressing structural and topological biases in offline HD maps caused by road renovation or sensor drift, PriorMaskMap predicts dense prior confidence to gate dual-prior query initialization and regulate BEV feature fusion, outperforming baselines with 81.4 mAP under corrupted priors.
Ray-Path-Aware Virtual Point Removal on 2D Layer-Wise Nearest Point Map: Addressing the erroneous virtual mirror points generated when LiDAR scans reflective materials (such as glass walls), this paper departs from explicit 3D surface modeling and heuristic symmetry assumptions by proposing an end-to-end removal framework in the 2D equirectangular projection domain using a Layer-Wise Nearest Point Map and Ray-Path-Aware Attention, improving the F1 score to 0.7895–0.8062 with over two orders of magnitude computational speedup.
Real-Time LiDAR Gaussian Splatting SLAM via Geometry-Aware Covariance Coupling: Addressing GPU memory explosion and error persistence in large-scale online LiDAR SLAM, this work establishes a bidirectional covariance coupling between fast G-ICP tracking and spherical 2D Gaussian Splatting mapping, delivering continuous high-fidelity surface reconstruction at >20 FPS real-time speed.
ReconDreamer-RL: Enhancing Reinforcement Learning via Diffusion-based Reconstruction: ReconDreamer-RL integrates video diffusion priors and kinematic physical constraints into 3D Gaussian Splatting scene reconstruction to build ReconSimulator, and couples it with a Dynamic Adversary Agent (DAA) and Cousin Trajectory Generator (CTG) to slash the closed-loop collision ratio of end-to-end autonomous driving to nearly one-third of prior RL baselines (and over 5x lower than imitation learning).
Reconstructing Dense Depth of Dark Scenes with Sparse LiDAR, Noisy Events, and Blurry RGB: Addressing the critical failure of motion-blurred long-exposure RGB imagery in low-light autonomous driving depth completion, ERU-Net tightly couples event-guided implicit neural feature restoration with confidence-aware depth upsampling, reconstructing sharp, temporally consistent dense depth sequences end-to-end.
Rethinking Training & Inference for Forecasting: Linking Winner-Take-All back to GMMs: This paper reveals that the widely used Winner-Take-All (WTA) loss in autonomous driving trajectory prediction is fundamentally a K-Means clustering objective rather than a Gaussian Mixture Model (GMM) likelihood. It is precisely this mismatch of "GMM modeling + WTA training" that leads to uninformative mode probabilities and over-segmentation. On this basis, the authors propose two training-free post-processing methods—test-time weighted merging and one-step EM fine-tuning—which significantly improve mode sorting quality and displacement accuracy.
RIGS: Radar-Informed Gaussian Splatting for Uncertainty-Aware 3D Occupancy and Motion Prediction: RIGS proposes a radar-informed 3D Gaussian splatting framework that systematically leverages 4D imaging radar range, RCS power, and Doppler velocity across scene representation, occupancy refinement, and motion estimation, enabling joint long-range 3D semantic occupancy, voxel-wise Dirichlet uncertainty, and dense 3D motion prediction.
Semantic-Aware, Physics-Informed, Geometry-Grounded Weather Synthesis: By disentangling weather synthesis into semantic anchoring, physics-informed anisotropic Gaussian particle simulation, and 3D geometry grounding, this work effectively elicits latent weather and particle priors from off-the-shelf video diffusion models to synthesize realistic, temporally coherent, and controllable adverse weather videos.
SGC-Lane: Monocular 3D Lane Detection with Standard-Definition Map Guidance and Lane Completion: To overcome the ill-posed nature of monocular 3D geometry reasoning and endpoint truncation in sparse-point representations, SGC-Lane adapts coarse Standard-Definition (SD) road-level centerlines into lane-level probability priors via front-view width prediction and introduces a step-consistency completion head (SCC-Head) with endpoint clustering, achieving a state-of-the-art 65.6% F1-score on OpenLane.
SkyLume: A Large-Scale Multi-Illumination Aerial Benchmark for Urban Scene Reconstruction and Beyond: Addressing the failure of photometric consistency caused by shifting illumination and moving cast shadows across multi-temporal aerial flights, SkyLume establishes the first large-scale real-world UAV benchmark comprising over 100K 6K five-directional oblique images across 10 urban regions, coupled with LiDAR ground truth and the Temporal Consistency Coefficient (TCC) to rigorously benchmark 3D reconstruction, novel view synthesis, and inverse rendering.
StreamOcc: Streaming Dense Voxel Representations for 3D Occupancy Prediction: StreamOcc integrates the streaming update paradigm with dense 3D voxel representations for occupancy prediction for the first time. It addresses the challenges of warping distortion and dynamic object feature degradation through rectified streaming voxel aggregation (StreamAgg) and query-guided aggregation (QueryAgg), achieving state-of-the-art (SOTA) performance on SurroundOcc and Occ3D-nuScenes with a real-time latency of 83.3ms/frame and a memory footprint of 2.8GB.
StreetForward: Perceiving Dynamic Street with Feedforward Causal Dynamics: StreetForward introduces a pose-free, tracker-free, and segmentation-free feedforward framework for 4D dynamic street reconstruction, leveraging causal masked attention to disentangle temporal directed motion alongside spatio-temporal rendering consistency to achieve high-fidelity spatio-temporal view synthesis and depth estimation.
SynHMR: Synergistic Joint-Mesh Modeling for LiDAR-based Human Mesh Reconstruction: Addressing the severe error propagation inherent to decoupled pipelines that estimate skeletal joints before reconstructing meshes under sparse LiDAR point clouds, SynHMR proposes a synergistic joint-mesh modeling framework using a Geometry-Aware Transformer with noise-augmented learning to enable bidirectional mutual refinement between skeleton and surface, achieving state-of-the-art accuracy across three benchmarks.
TEX-Drive: Temporal Perception Meets Experience-Guided Mixture-of-Experts for End-to-End Autonomous Driving: Addressing the disjointed modeling of temporal perception and functional specialization in end-to-end autonomous driving, TEX-Drive integrates an explicit key-frame temporal perception unit with an experience-driven Mixture-of-Experts planner, achieving a state-of-the-art Driving Score of 68.22 and 40.62% Success Rate on the closed-loop Bench2Drive benchmark.
The Language of Visual Attention: Modeling Scanpaths via Autoregressive Token Prediction: Reframes eye movement scanpath prediction purely as a discrete coordinate token generation task via autoregressive Large Vision-Language Models (LVLMs), replacing rigid hand-crafted geometric priors, setting a new state of the art of 2.18 bits Information Gain on MIT1003 (+46% over feature-matched DeepGaze III), and natively enabling prompt-conditioned interventions across durations, subjects, and search goals.
TriO: Tri-Modal Unsupervised Occupancy World Model for Anything Perception: TriO establishes an unsupervised 4D continuous occupancy world model leveraging camera, LiDAR, and RADAR as both inputs and self-supervisory sources, achieving robust zero-shot road obstacle segmentation and flow forecasting without human annotations.
Uncertainty-Driven Gaussian Sphere Propagation for 3D Semantic Segmentation: To address the breakdown of deterministic predictions and isotropic feature aggregation across physical occlusions and sparse boundaries in 3D point clouds, this paper presents Uncertainty-driven Gaussian Sphere Propagation (UGSP), which constructs continuous topological paths between Monte Carlo Dropout-partitioned uncertainty regions and injects reliable anchor semantics into geometric voids via Spherical Harmonics-based anisotropic Gaussian spheres.
UniFlow: Zero-Shot LiDAR Scene Flow for Autonomous Driving: UniFlow overturns the long-standing belief from 3D detection and segmentation that multi-dataset LiDAR training degrades performance, proving that class-agnostic scene flow models transfer robustly across sensors when trained on unified datasets with temporal alignment and minimal geometric augmentations, achieving state-of-the-art results on Waymo and nuScenes while outperforming dataset-specific models on unseen TruckScenes and AEVAScenes benchmarks.
Unpaired Geometry-Guided Sim2Real Translation for Autonomous Driving: To tackle the lack of paired geometric supervision and the prohibitive parameter overhead of multi-condition control in autonomous driving Sim2Real transfer, this paper proposes Cross-Domain Control Transfer (CDCT), which injects domain-agnostic geometry guidance derived from synthetic score differences into a real-domain appearance prior, and designs the Lightweight Multi-Condition Adapter (LMCA) to achieve high-fidelity unpaired translation and substantially improve downstream perception.
VectorReLoc: Reliable Vectorized SD Map Visual Re-localization with Contrastive Feature Alignment: VectorReLoc discards computation-heavy dense rasterized BEV matching pipelines and proposes the first sparse re-localization framework operating directly on vectorized map representations, combining contrastive feature alignment with a per-dimension reliability predictor to achieve centimeter-level 3-DoF pose correction in 2–5 ms on edge computing chips.
VVSim: A Large-Scale Aerial-Ground Dataset and Benchmark for Cooperative Perception: Addressing the view constraints and occlusion vulnerabilities of conventional ground-based cooperative perception, this paper presents VVSim, the first large-scale aerial-ground cooperative perception dataset comprising 61k time-synchronized multimodal frames, alongside VVFormer, a novel framework leveraging cross-view voxel projection, height-aware attention, and agent-aware deformable cross-attention to advance cooperative 3D perception.
WarpI2I: Image Warping for Image-to-Image Translation: This paper proposes a model-agnostic, parameter-free warp-unwarp framework that spatially enlarges salient regions before encoding and subsequently reverses the warping, preserving fine-grained structural information in LDM image translation without increasing the latent space resolution. Meanwhile, a lightweight FLUX-based synthetic data pipeline is proposed, yielding significant improvements with minimal overhead for human/driving scene relighting and weather/time-of-day driving translation tasks.
Who Does What and Where to Go: Orthogonal Alignment and Hierarchical Planning for Multi-Entity Trajectories: To tackle role-action confusion and physical interpenetration in multi-entity text-driven trajectory generation, ProgTraj-Director orthogonalizes static visual masks and dynamic action priors via factorized tensor fusion (StructVLA), and deploys a geometry-aware Bird’s-Eye View (BEV) blueprint to guide progressive 3D diffusion denoising, cutting the collision rate to 7.3% and securing 96% action attribution accuracy on the newly introduced Stepwise-ME dataset.
ZTRS: Zero-Human Demonstration End-to-end Autonomous Driving with Trajectory Scorer: ZTRS exhaustively evaluates rule-based rewards over a fixed trajectory vocabulary and replaces expert trajectory imitation and sparse action sampling with Exhaustive Policy Optimization (EPO), reaching 45.5 EPDMS on Navhard with V2-99; “zero-human demonstration” means that training uses no per-sample expert imitation targets, not that the trajectory vocabulary and pretrained visual backbone contain no data-derived priors.