augmented reality developers

Abstract: Perceiving and remembering the visual world is fundamental to navigating and interacting with our environment. Current 4D foundation models, such as camera-controllable video models or 4D reconstruction models, can perceive and reconstruct dynamic environments, but how well they remember what they have perceived remains an open question. Existing benchmarks largely rely on pixel-level metrics and lack ground truth for objects once they leave the field of view, making them unable to evaluate visual memory in an object-centric manner against references. To fill this gap, we introduce PersistBench, a dataset and metric suite that leverages 360° videos as omniscient ground truth and proposes three evaluation aspects: object permanence, motion continuity, and appearance preservation. Evaluating various models across diverse categories reveals that current models can only maintain short-term consistency that degrades significantly once objects leave the field of view. Our findings highlight the gap between current model capabilities and robust visual memory ("seeing is not remembering"), providing guidance for future development of 4D foundation models. Dataset and code are available on the project page: https://guangzhaohe.com/persistbench.

Abstract: Tactile signals provide direct contact and force measurements that are essential for understanding physical interactions and enabling dexterous robotic manipulation. However, tactile sensing requires direct measurement at contact interfaces, making large-scale data collection reliant on intrusive, costly, and restrictive instrumentation. We present TouchSight, a monocular egocentric vision framework for dense full-hand contact force prediction that leverages 500 hours of pressure-glove recordings and extensive hand-object interaction (HOI) data. To address the appearance gap between gloved training data and bare-hand real-world scenarios, we construct TwinTouch-20H: 20 hours of paired visual data in which generative video models re-render gloved recordings as bare-hand observations against new backgrounds while preserving the original measured tactile labels. TouchSight predicts dense force from both gloved and generated bare-hand videos, outperforms prior contact prediction methods on OakInk2, qualitatively generalizes to natural bare-hand egocentric videos from unseen datasets, and improves consistently as glove supervision scales. These results demonstrate that dense tactile signals can be recovered from egocentric vision alone, without tactile instrumentation at capture time.

Abstract: Feed-forward 3D reconstruction provides an efficient paradigm for scene modeling from image sequences. Scaling these models to large monocular scenarios are constrained by excessive GPU memory footprint, degraded local geometry, and long-term trajectory drift. Existing chunk-based optimization strategies provide limited geometric constraints and fail to maintain global consistency over extended trajectories. We present a unified framework for stable and scalable feed-forward 3D reconstruction from long monocular sequences. Our approach builds on coarse-to-fine trajectory alignment augmented by lightweight geometric prior injection. Distilling monocular geometric cues into the feed-forward backbone via LoRA adaptation improves depth accuracy on fine structures while preserving inference efficiency. We introduce a hybrid-weight sparse ray-field optimization that leverages high-frequency geometric features to guide local point-cloud refinement and enforce consistent inter-frame ray constraints. Unlike prior chunk-based methods, this establishes strong cross-frame geometric coupling while maintaining scalability. Finally, an efficient trajectory stitching strategy with joint ray-error optimization explicitly reduces accumulated drift. Extensive experiments show that our approach achieves competitive trajectory accuracy compared with representative SLAM systems, while maintaining globally consistent 3D reconstruction in large-scale scenarios.

Abstract: Pre-trained 3D vision models have substantially advanced point cloud analysis, yet adapting them to downstream tasks via full fine-tuning is computationally expensive and storage-intensive. Parameter-Efficient Fine-Tuning (PEFT) offers a promising alternative by reducing both adaptation cost and storage burden. However, existing prompting-based approaches ignore the intrinsic geometric structures of point clouds, thereby limiting their adaptation capability. This limitation stems from their inability to encode both fine-grained geometric cues and coarse-grained structural semantics, as well as failing to propagate such information effectively through the model hierarchy. To address these challenges, we propose GAPrompt++, a multi-granular geometry-aware prompting method that provides richer geometric guidance for efficient 3D task adaptation. Specifically, we introduce a Point Shift Prompter that extracts multi-granular geometric features across different scales, enabling instance-specific geometric adjustments during adaptation. Next, a Keypoint Prompter adaptively generates point-level prompts to highlight local geometric saliency and fine-grained structural details. Furthermore, a Prompt Propagation mechanism injects these multi-granular geometric cues throughout the feature extraction hierarchy, strengthening the ability to capture essential geometric characteristics. Extensive experiments show that GAPrompt++ achieves state-of-the-art performance among prompting-based PEFT methods and even surpasses full fine-tuning across diverse benchmarks, while requiring less than 2\% trainable parameters. In addition, to address the saturation of existing evaluation datasets, we construct two more challenging benchmarks derived from 3D Gaussian Splatting and Multi-View Stereo reconstruction, offering diverse and realistic point cloud scenarios to promote future research.

Abstract: Training-free weighted aggregation is widely used to lift 2D semantic features onto 3D Gaussians for open-vocabulary scene understanding, yet its theoretical role remains insufficiently understood. Existing analyses typically justify this operation from the rendering side, treating Gaussian features as linearly composable Euclidean variables for reconstructing 2D feature maps. However, this view does not match downstream 3D usage, where each Gaussian is often queried independently in a cosine-based embedding space. We revisit feature lifting from the 3D side and formulate per-Gaussian assignment as a cosine alignment problem on the CLIP unit sphere. Under this objective, the L2-normalized semantic back-projected feature emerges as the closed-form solution, providing a complementary interpretation of the standard lifting rule from the perspective of per-Gaussian semantic assignment. The same formulation further yields a norm decomposition into intra-view and inter-view consistency, suggesting that feature magnitude itself can serve as a semantic reliability signal. Calibrated by effective multi-view support, this reliability score guides a mode-voting refinement that preserves CLIP feature validity by avoiding linear averaging. Experiments on open-vocabulary 3D semantic segmentation show that NormLift is an efficient, training-free framework that achieves strong performance across evaluation protocols.

Abstract: Vision-language navigation requires aligning language with visual observations while maintaining spatial understanding over time. Geometry foundation models (GFMs) expose intermediate representations throughout their hierarchy, but how navigation policies should use these features and retain historical geometric evidence remains unresolved. We introduce \method{}, a streaming VLN framework that addresses these questions across \textbf{representation depth} and \textbf{navigation time}. Hierarchical GFM--VLM fusion couples earlier, intermediate, and later GFM representations to successive policy stages instead of repeatedly injecting a terminal feature. Navigation-aware GFM memory retains historical VGGT global-attention KV states according to instruction relevance, geometric confidence, and transition novelty under a bounded per-layer budget. Retained states provide geometric context for subsequent observations before fusion with the policy. Across R2R-CE and RxR-CE, \method{} achieves strong performance using a single RGB stream without additional navigation-specific external data. Controlled ablations show that multi-depth coupling substantially outperforms repeated terminal-feature injection at matched fusion locations. Bounded navigation-aware retention preserves navigation performance while considerably reducing GFM-KV memory relative to larger-memory temporal retention. These findings support jointly examining the geometric representations exposed to the policy and the historical evidence retained for future inference. Code will be released upon acceptance at https://humanoid-research.github.io/adageovln/.

Abstract: We present CADSplat, a framework that reconstructs photorealistic, geometrically accurate digital twins from sparse ($<15$ views), wide-baseline posed images of an object by regularizing 3D Gaussian Splatting (3DGS) with an explicit CAD shape prior. Using such a prior requires finding a CAD model whose shape resembles the object depicted in the images and determining the pose of each camera relative to the object. We obtain both by matching segmented object silhouettes against silhouettes rendered from a CAD library and keeping the camera-to-object poses of the best-matching model. We then anchor 3D Gaussian primitives to the surface of the retrieved model and jointly optimize the 3DGS parameters, the camera-to-object registration, and a non-rigid deformation field to account for shape differences between the physical object and the CAD model. Across two real-world datasets, CADSplat outperforms unconstrained, few-shot, and mesh-texturing baselines and degrades gracefully to as few as 3 views. Our experiments show that most of the gain in rendering quality comes from how the splats are constrained---a fixed set of splats tied to a surface and moved by a single smooth deformation field---rather than from the CAD shape itself. The CAD model adds shape knowledge where views are scarcest, in the sparsest captures and on strongly self-occluded objects, and it places every camera in the object's own frame. This enables applications beyond novel-view synthesis, such as markerless augmented reality registration, per-image object pose estimation, physical simulations, and the transfer of part labels from the design to the reconstruction.

Abstract: Feed-forward 3D foundation models such as VGGT predict cameras, depth, and point maps in a single pass, but can fail silently under low overlap, low parallax, and extreme relative rotation. Stratified analyses over these factors show that these failures are governed by geometric conditioning and are poorly captured by native aleatoric confidence. We introduce GeoCond, a lightweight reliability adapter for frozen feed-forward 3D backbones. GeoCond reads the backbone's predicted geometry and outputs pose-level uncertainty and a refinement gate. During training, it can be supervised by frame-permutation orbit variance, ground-truth pose error when labels are available, or cycle residuals from unlabelled independent pose graphs. At inference, the default head requires only one backbone pass and a small MLP. On VGGT, GeoCond improves out-of-distribution (OOD) AUSE (area under the sparsification-error curve; lower is better) from $0.32$ to $0.20$ over native confidence, transfers zero-shot to outdoor extreme-view scenes, and avoids the collapse caused by applying bundle adjustment uniformly. Across multiple backbones, cycle-distilled variants provide a ground-truth-free adaptation route, including cases where permutation variance vanishes on equivariant models. The same reliability signal supports gated refinement, pose-graph weighting, calibration, curation, and capture decisions. Reliable feed-forward 3D reconstruction requires not only predicting geometry, but also knowing when that geometry should be trusted.

Abstract: Novel view synthesis from unposed multi-view images remains challenging, as the model must jointly learn scene representations and camera parameters without pose supervision. Existing approaches largely fall into two extremes: implicit latent-space rendering is flexible and easy to optimize, but often yields weakly grounded camera estimation; explicit 3D representations provide stronger geometric grounding, but introduce heavier parameterization and more fragile optimization. In this paper, we present IRIS, a fully self-supervised framework that provides a practical middle ground between these two paradigms. Instead of decoding free latent tokens or reconstructing fully explicit 3D primitives, IRIS represents the scene as a latent neural field and renders novel views by querying this field under self-predicted cameras. Specifically, projected features from reference views are aggregated at sampled 3D points to form point-wise latent features, which are then composed along target rays for rendering. This design preserves the flexibility and optimization stability of implicit modeling, while introducing stronger geometric structure than unconstrained latent rendering. Extensive experiments show that IRIS achieves strong novel view synthesis quality with competitive pose accuracy under fully self-supervised learning. Our project page: https://leo-frank.github.io/IRIS

Abstract: Semantic occupancy prediction gives an embodied agent a voxel-level account of where space is free, occupied, and semantically meaningful. In RGB-D pipelines such as EmbodiedScan, the image encoder is often left as a default module, even though its features are the visual evidence later sampled into the 3D grid. We study this design choice directly. A central finding is that changing the 2D backbone improves occupancy accuracy more than several carefully designed occupancy architectures or modules. We keep the main RGB-D projection, depth branch, and occupancy head fixed, and replace only the image backbone. The compared encoders are CLIP-ResNet, CLIP-ViT, BLIP2, and DINOv2. Under the controlled setting, the measured mIoU changes substantially: DINOv2 obtains 30.55\%, BLIP2 obtains 29.49\%, CLIP-ViT obtains 24.33\%, and CLIP-ResNet obtains 17.41\%. The stronger encoders also exceed the original EmbodiedScan ResNet-50 baseline without modifying the downstream 3D fusion pipeline. Class-level results give a more detailed picture: DINOv2 is stronger on many layout and structural categories, whereas BLIP2 remains close on several object-centered classes. CLIP-ViT improves clearly over CLIP-ResNet, showing that the way CLIP features are exposed as dense tokens matters for voxel lifting. These results indicate that the image backbone is not a secondary engineering detail in embodied semantic occupancy, but a major source of variation in the final 3D prediction.

Abstract: Streaming 3D reconstruction demands both speed and temporal fidelity, goals that existing methods undermine by updating every Gaussian every frame, even in static regions. We present DecoGS, a method for efficient online training of 3D Gaussians from streaming videos. Unlike prior methods that update the entire scene indiscriminately, DecoGS introduces an adaptive mechanism that selectively focuses optimization on spatiotemporal regions exhibiting motion or photometric changes. This targeted training strategy eliminates redundant updates that cause flickering and drift in nominally static regions, while enabling fast, high-fidelity scene updates. The pipeline further integrates region-aware Gaussian management through gradient gating and efficient visibility filtering to maintain temporal coherence and a compact memory footprint. On N3DV and MeetRoom, DecoGS achieves 34.55 and 31.60 dB PSNR respectively, outperforming all streaming and offline baselines, while rendering at 261 FPS with $70\times$ lower temporal flicker than the best prior method, requiring no large-scale pretraining.

Abstract: Beyond model inference, the decoding stage, which converts raw network outputs into task-level representations, constitutes a significant portion of the execution cost. Despite its practical impact, prediction decoding has received comparatively little attention and is often implemented using generic CPU routines or inefficient GPU kernels, limiting the benefits of advances in model efficiency. In this work, we investigate the decoding overhead associated with a recent sinusoidal centroid encoding for Instance Segmentation, in which each pixel regresses a positional embedding of its instance centroid. This approach allows flexible segmentation without predefined proposals, but extracting instance masks from dense embeddings incurs a high computational cost. We present an optimized CUDA-based implementation of the decoding algorithm tailored to this encoding, explicitly addressing challenges related to parallelization, synchronization, and memory access on modern GPUs. Our solution significantly reduces decoding overhead and improves End-to-End inference latency, outperforming both CPU-based approaches and naive GPU implementations. The results demonstrate that efficient decoding is essential to fully exploit the advantages of advanced output representations and highlight the importance of jointly designing encoding schemes and their decoding algorithms for real-time computer vision systems.

Abstract: Category-level object pose estimation seeks to recover a similarity transform $(R,t,s)$ for unseen instances without instance-specific CAD models. Most competitive methods are correspondence-based: prior-free variants regress canonical (NOCS) coordinates directly from local observations and implicitly memorize the canonical frame in the weights, which ties the parameters to category-typical orientations and hurts generalization under distribution shift; prior-based variants introduce a category prior but typically follow a serial deform-then-align pipeline, where underconstrained canonical completion can corrupt correspondences and induce error cascades in pose. We propose PriorPose, a reference-guided correspondence framework that keeps the category prior explicit and solves canonicalization and alignment jointly in a shared feature space. A reference-guided seeded transformer embeds the partial observation and the category prior as token sets and fuses them via geometry-aware seeds, from which the network jointly predicts a per-point NOCS field for visible points and a canonical deformation of the prior that reconstructs a full canonical instance, while a deep pose head regresses $(R,t,s)$ from the induced correspondences. A two-part shape consistency objective, with canonical-space and camera-space consistency losses, couples correspondence, deformation, and pose, reducing reliance on memorized canonical orientations and avoiding deform-then-align error cascades. Experiments on standard and larger-category benchmarks demonstrate that PriorPose sets new state-of-the-art results on most evaluated metrics, especially under strict pose thresholds, while remaining competitive on relaxed pose and IoU metrics and showing improved robustness under shape variation and domain shift.

Abstract: Learning manipulation from human video requires high-fidelity hand-motion reconstruction in metric units. Today's metric hand labels come from studio rigs and instrumented headsets, and both are confined in the same two ways: neither leaves a prepared setting, and neither is checked against an independent reference. Unconstrained head-worn recording promises the opposite trade-off, scaling with the number of people wearing a device. We therefore introduce MEgoVista, an offline pipeline that turns a single unprepared MEgo View recording into metric two-hand and head motion in one gravity-aligned world frame. Three properties set it apart from existing egocentric reconstruction systems: first, it reconstructs in settings studio volumes and tabletop rigs cannot reach, settling hand ownership at detection so bystander hands stay out of the wearer's trajectory; second, it takes its metric gauge from calibrated stereo rather than a monocular prior, installing scale at initialisation so policies receive physical units, not arbitrary coordinates; third, both outputs are scored inside a motion-capture volume against independent Chingmu optical capture, under a protocol that audits its own reference and charges what a method declines to predict. MEgoVista is offered as a measured route from egocentric video to metric hand supervision, one that widens where such labels can be gathered.

Abstract: Virtual try-on (VTON) enables customers to visualize how fashion products appear when worn and has become an important technology for online shopping. While recent advances have substantially improved garment VTON, jewelry remains a challenging and underexplored category due to its small size, rigid structure, and sensitivity to fine-grained visual details. Realistic jewelry VTON requires not only faithful appearance transfer but also accurate scale and placement relative to the wearer. Existing jewelry VTON methods typically rely on mask guidance, whereas mask-free approaches lack explicit guidance for modeling the product scale. To bridge this gap, we introduce JVTO-Bench, a benchmark dataset for scale-faithful jewelry VTON, providing reference source target triplets with real-world product-scale annotations across four major jewelry categories. Building upon this benchmark, we propose JewelTry, a mask-free diffusion framework for scale-aware jewelry VTON. JewelTry incorporates a scale adapter that encodes product dimensions into a scale token, enabling the model to learn scale relationships between jewelry items and surrounding human anatomy in-context. To further improve jewelry consistency, we introduce a single-directional condition attention mechanism and an attention refinement loss that preserve both coarse geometry and fine-grained structural details of the reference jewelry. Extensive experiments show that JewelTry achieves a balance among visual fidelity, background preservation, object consistency and scale accuracy, establishing a strong baseline for mask-free, scale-aware jewelry virtual try-on.

Abstract: Vision-language models excel at 2D image understanding but remain limited in 3D spatial reasoning. Progress is hindered by limitations in current benchmarks. First, 3D datasets often rely on point clouds that capture geometry but discard rich visual features like texture, text, and materials. Second, annotations treat objects in isolation while ignoring real-world hierarchical organization (scenes, rooms, functional areas, object groups). Third, evaluation tasks focus narrowly on basic recognition rather than multi-step spatial reasoning. In this context, we introduce SceneBench, a benchmark of 966 photorealistic 3D scenes reconstructed with Gaussian Splatting and densely annotated with hierarchical semantics spanning scenes, rooms, functional areas, object groups, and individual objects. These annotations are produced through a human-in-the-loop pipeline combining vision-language models with roughly 1,500 human-hours of iterative refinement and verification, producing over 183K annotated nodes with textual descriptions and 3D bounding boxes. Building on this representation, we define three evaluation tasks: Existence-Based Questions probing object attributes, Spatial Intelligence Questions covering counting, size comparison, distance, and directional relations, and Grounded Question-Reasoning-Answer (QRA) triplets requiring multi-step reasoning across semantic levels. Experiments with state-of-the-art vision-language models show that while models perform well on basic recognition tasks (e.g., up to 85% accuracy for detection), performance drops substantially on hierarchical and compositional reasoning (e.g., down to 60% for counting), revealing limitations not captured by existing benchmarks. SceneBench provides a realistic testbed for developing and evaluating models capable of fine-grained spatial reasoning in photorealistic 3D environments.

Abstract: Part-level control is essential for modern 3D asset creation, where objects are frequently edited, reused, animated, or fabricated through their individual components. In many such workflows, users need only several specific components rather than a complete object decomposition. However, existing 3D generation methods produce all parts regardless of user intent, while promptable 3D segmentation methods typically output partial surfaces instead of reusable complete meshes. In addition, image-conditioned part generators further struggle to preserve hidden geometry and accurate placement without directly conditioning on the source mesh. To address these problems, we present SAM3D-Part, a prompt-driven framework for selective part generation from input 3D object meshes. Given a source mesh and a part prompt, SAM3D-Part first encodes the source geometry into compact mesh features and aligns them with the rendered image, selective mask, and point-map observations via pixel-wise channel fusion. The fused representation conditions a feed-forward generative model to produce only the queried component as a completed mesh. To place the generated part back into the source coordinate frame, SAM3D-Part predicts dense per-voxel correspondences and estimates the part transformation from distributed spatial evidence rather than a single global pose code. For sequential multi-part queries, previously generated parts are stored in a part cache and reused as contextual constraints, reducing conflicts among independently requested components. Extensive experiments and ablations demonstrate that SAM3D-Part can significantly improve source alignment, reduce conditioning cost, and enable consistent selective part generation, achieving state-of-the-art. Code and weights will be available at https://github.com/Jiahao620/sam3d-part.

Abstract: Diffusion models generate images through an iterative diffusion process, and recent studies have demonstrated that the intermediate feature maps produced during this process contain rich visual representations, leading to their adoption across a variety of downstream tasks. However, most existing approaches are limited to either using a single feature map at a specific timestep or aggregating feature maps across multiple timesteps. We observe that intermediate representations in the diffusion process form meaningful trajectories along the time axis. In particular, the representation of each spatial patch evolves progressively throughout the generative process, encoding semantics that are difficult to capture from static snapshots alone. This observation motivates the need to treat diffusion representations as temporally structured trajectories rather than static snapshots. To this end, we propose Diffusion Trajectory Modeling (DTM), a framework that interprets the temporal evolution of each spatial patch as a trajectory and leverages it for semantic correspondence. By effectively modeling patch-wise trajectories generated across multiple timesteps, DTM captures correspondence cues that prior methods are not designed to capture. We further demonstrate empirically that spatially corresponding patches form similar trajectory patterns throughout the diffusion process, suggesting that the temporal axis of diffusion carries semantic information. Experiments on SPair-71k, SPair-U and AP-10K show that DTM achieves strong performance, presenting a new perspective for exploiting diffusion representations from a trajectory-centric viewpoint.

Abstract: Multi-model fitting involves fitting multiple models accurately in a noisy environment. It is the basis for computer vision tasks such as scene reconstruction and mixed reality. However, its performance is often limited by insufficient feature utilization, inefficient optimization, model overlap, and the non-differentiable pipelines. To overcome these limitations, we introduce a robust coarse-to-fine framework called Learning Neighbor Regions (LNR). Recognizing that substantial computational resources are wasted on numerous bad minimum sets, we propose the coarse-level module. This module utilizes a neural network to extract and analyze geometric feature of both local point-wise relationships and global contextual information in minimum sets, outputting confidence to pre-select a small number of good minimum sets, thereby enhancing overall efficiency before solving hypotheses. To address model overlap, LNR encodes neighbor region features for each hypothesis in its fine-level module. These region features consist of geometric features of neighboring data points, which can be used by multiple regions simultaneously. This design allows the neural network to individually refine and score each hypothesis. Importantly, LNR is trained to learn directly from data point features rather than from the hypothesis parameters, thus avoiding differentiating the sampling process and the model solvers. Extensive experiments on four classic multi-model fitting tasks demonstrate that LNR achieves state-of-the-art performance. The analysis suggests that LNR can be easily adapted to various robust multi-model fitting tasks.

Abstract: 3D Gaussian language fields provide an explicit, spatially grounded representation for 3D visual question answering (VQA), but their dense semantic features can require tens of thousands of embeddings per scene, resulting in substantial storage, memory, and inference costs. We investigate how much of this representation is actually necessary for downstream reasoning. Starting from a full embedding representation, we systematically sparsify its semantic embeddings, including the previously underexplored regime below a single image-equivalent block down to 8 visual tokens. We compare random, geometric, semantic, and joint spatial-semantic selection strategies and introduce an object-based sparsification method that distributes the token budget across detected object instances while retaining background context. Experiments on ScanQA and MV-ScanQA reveal substantial redundancy in dense Gaussian language fields. Strong VQA performance is retained with only a few hundred semantic embeddings, corresponding to less than 1% of the original representation. Object-based selection performs well relative to others, with only modest observed changes down to 256 tokens. At this budget, SparseTalk retains 0.80% of SplatTalk's 32,076-token inference input and 0.332% of the mean 77,207-Gaussian dense field, increasing inference throughput while reducing decoded-feature memory 125-fold.

Abstract: Continuous-environment vision-and-language navigation (VLN-CE) requires interpreting natural-language instructions in unseen 3D environments and executing continuous low-level actions. Existing methods often depend on LiDAR, panoramic cameras, or extra sensors; separate geometric-mapping and semantic-navigation visual representations can cause long-trajectory spatial-semantic inconsistencies. We propose LG-VLN, a monocular zero-shot framework with shared visual features and LangGraph-based state orchestration. An online feed-forward 3D reconstruction network predicts depth, camera poses, and dense point clouds for agent-pose estimation and global map fusion. Geometry and navigation share dense CleanDIFT features: semantic consistency rejects incorrect inter-frame correspondences, while target-instance constraints define visual references whose similarity combines with local BLIP-2 image-text relevance to form a semantic value map. LangGraph represents instruction parsing, geometric perception, semantic value updates, path planning, action execution, and failure recovery as a directed state graph with conditional transitions, persistent state, and modular recovery mechanisms. On a fixed 550-episode subset of the R2R-CE val-unseen split, LG-VLN achieves 21.3% success and 12.1% success weighted by path length. Ablations show shared semantic features improve navigation, further boosted by combining visual similarity and image-text relevance. Results establish shared visual representations and explicit state orchestration as effective for zero-shot VLN-CE using monocular RGB alone. Code will be publicly released for reproducibility.

Abstract: We study relative position encoding for multi-view vision Transformers under camera heterogeneity, including varying fields of view (FoVs) or projection models. Existing rotary relative position encodings commonly use image-plane positional coordinates, producing projection-dependent relative phases and inconsistent geometric cues for cross-projection attention. We introduce G-ray, a ray-level relative position encoding whose rotary phases are parameterized by camera-local ray angles. The same camera-local ray pair induces the same relative phase across projections, providing projection-invariant positional consistency. G-ray can be used directly or integrated with existing encodings, retaining complementary geometric cues without additional learned parameters. We validate G-ray in three host encodings, RoPE, GTA, and RayRoPE, across 3D reconstruction and novel-view synthesis (NVS). Across three heterogeneous 3D reconstruction benchmarks at 50 views, G-ray leads all six averaged metrics and reduces mean pointmap relative error by 45.8% over MapAnything, with calibration supplied to both. Trained exclusively on homogeneous pinhole images, the 3D reconstruction model handles mixed pinhole and non-pinhole inputs without retraining and remains competitive on homogeneous pinhole 3D reconstruction protocols. For NVS, GTA and RayRoPE improve with G-ray under joint viewpoint and FoV variation. The project's webpage is available at https://g-ray-project.github.io/.

Abstract: We present PhysBrain 1.5, a unified model for understanding physical environments, generating actions, and predicting future states. Motivated by the physical loop of observation, interaction, and environmental change, we bring these capabilities into a common learning framework. Starting from a general vision--language model, we encode language responses, end-effector motion, and dense visual targets as discrete sequences and jointly optimize them with autoregressive next-token prediction. Pre-training draws its embodied supervision entirely from human interaction videos, using task-centered episodes to pair semantic and spatial context with recovered motion and subsequent observations. We then adapt the model through supervised fine-tuning on a mixture of human demonstrations, robot trajectories, and simulated experience. Across 28 embodied understanding benchmarks, our 8B model achieves an average score of 72.5, setting a new open-source state of the art and performing on par with leading proprietary models such as GPT-6-Astra and Gemini 3.6 Flash. It achieves the best open-source results on 14 benchmarks while retaining general multimodal capabilities. Beyond these understanding evaluations, qualitative examples show the model's ability to produce end-effector trajectories and predict future scenes through spatially aligned RGB, depth, and robot-mask outputs.

Abstract: Recently, 3D Gaussian Splatting SLAM (3DGS-SLAM) has gained significant momentum in simultaneous localization and 3DGS scene reconstruction. In real-world scenarios with rapid camera motion and cluttered dynamic environments, existing methods rely on the stability of the underlying scene reconstruction to model uncertainty. This leads to a circular dependency between camera tracking accuracy and reconstruction quality: reconstruction instabilities degrade uncertainty modeling, which affects accurate camera tracking and static scene reconstruction. To address this, the paper proposes SCOUT-SLAM, a structurally-coupled dual-uncertainty framework in which both uncertainties are estimated from a shared base network. A low-rank adaptation of this network, trained on multi-view feature consistency, estimates a tracking uncertainty that does not depend solely on the reconstruction quality. A spatially-adaptive prior modulates the network's training objective so that reconstruction instability does not inflate uncertainty on static regions, keeping the shared representation intact for both branches. Evaluations on dynamic benchmarks (TUM RGB-D, Bonn Dynamic, Wild-SLAM MoCap) demonstrate that SCOUT-SLAM achieves state-of-the-art camera tracking accuracy and artifact-free static scene reconstruction.

Abstract: We study how to update a pre-built aerial scene with a newly captured, unposed ground-view sequence. The aerial scene already contains a reliable metric Structure-from-Motion (SfM) reconstruction and a pre-trained 3D Gaussian Splatting (3DGS) model, whereas the ground-view sequence is collected later to add street-level appearance but has unknown camera poses and global scale. Registering this sequence to the aerial SfM reconstruction is challenging because single-image cross-view localization is brittle and long trajectories are prone to drift. To address these challenges, we present SkyAnchor, which treats the existing aerial scene as a fixed scaffold for ground-view registration and scene update instead of jointly reconstructing aerial and ground imagery from scratch. It first localizes short groups of consecutive ground frames against geometrically verified aerial support, producing sparse anchor poses. It then recovers the full ground trajectory with anchor-constrained submaps, fixing the front and rear anchor poses during incremental registration and bundle adjustment. Finally, it inserts filtered ground Gaussians while preserving the aerial view, followed by lightweight joint refinement. Experiments on seven real aerial--ground scenes show accurate metric ground trajectories and updated 3D Gaussian scenes with strong aerial- and ground-view rendering quality.

Abstract: Learned components are increasingly integrated into geometric visual--inertial estimators to provide motion, depth, bias, uncertainty, or confidence cues. Yet it remains unclear whether gains arise from useful learned priors or from changes in the backend, calibration, initialization, temporal association, or evaluation gauge. We present a controlled framework for learning-augmented visual--inertial estimation that separates fusion gain from the incremental value of a learned prior and evaluates four evidence layers: local motion consistency, global trajectory accuracy, physical-state correctness, and numerical consistency. We instantiate the framework with a MonoViT-based monocular motion prior added as a local relative-motion factor to an unchanged VINS backend. We compare Original VINS and learned-prior VINS under matched sensor streams, timestamps, initialization, frontend/backend settings, and camera--IMU extrinsics, while probing calibration, initialization, state coupling, scale, bundle adjustment, and loop closure. On KITTI, with fixed reference extrinsics, translation APE RMSE is 31.4 m for Original VINS and 31.8 m with the learned prior. Across four recordings, the prior changes mean APE by only -0.2%, while a five-times-higher weight worsens it by 8.2%. Online extrinsic updates increase mean APE by 45.7% and 52.1%, respectively, while mean RPE changes by less than 2%. These results show that fusion performance alone cannot establish the value of learned priors. Reliable evaluation requires same-backend controls and joint analysis of prior compatibility, calibration, initialization, global drift, physical-state error, and backend consistency.

Abstract: Most existing visual odometry (VO) systems treat feature correspondences as deterministic measurements or assign uniform uncertainty, ignoring the inherent localization ambiguity of different observations. However, correspondence uncertainty is often anisotropic due to image structures such as edges, repetitive patterns, and motion blur, which can significantly affect geometric optimization. In this work, we propose MomentBA, a geometry-aware bundle adjustment framework that derives anisotropic correspondence uncertainty from second-order spatial moments of local similarity responses. Instead of introducing additional covariance prediction networks, the proposed method directly converts matching response distributions into interpretable covariance estimates and incorporates them into bundle adjustment as correspondence-specific information matrices for uncertainty-aware residual weighting. Furthermore, the proposed formulation is integrated into a differentiable optimization framework, establishing a direct connection between correspondence uncertainty and geometric estimation. Experiments on the EuRoC MAV and TartanAir v1 Hard datasets demonstrate that MomentBA improves monocular visual odometry accuracy compared with existing feature-based and learning-based approaches. The proposed anisotropic covariance model achieves lower rotational errors and more robust trajectory estimation than fixed and isotropic uncertainty models, validating the effectiveness of geometry-induced uncertainty modeling for challenging visual environments.

Abstract: Fine-grained robotic manipulation depends on understanding parts, not only whole objects. Existing 3D foundation models tend to be either generalized but object-aware, or part-aware but limited to closed-set taxonomies, which weakens zero-shot transfer. We study text-conditioned 3D part segmentation, where a free-form phrase selects a functional part on point cloud. We introduce UniPart, a feed-forward cross-modal 3D Transformer that conditions CLIP text embedding. To scale supervision, we build LangPart-1M with 160K+ Objaverse assets and 8M text to part pairs using multi-view consistent part generation. We further manually label a high-quality subset, LangPart-4K, for fine-tuning and evaluation. UniPart achieves strong zero-shot results on open-vocabulary part benchmarks and transfers to language-conditioned part grasping in real world.

Abstract: Multimodal AI, powered by Large Language Models (LLMs) and Vision-Language Models (VLMs), is transforming assistive technologies by enabling simultaneous processing of visual and textual data. This advancement holds significant promise for over 43 million visually impaired and neuro-divergent individuals worldwide who face persistent challenges in navigating indoor and outdoor environments due to limited spatial awareness and insufficient environmental cues. Existing navigation aids often lack comprehensive 3D scene understanding, relying on constrained route-based strategies that hinder user autonomy. In this paper, we introduce a novel end-to-end framework that integrates LLMs, VLMs and digital twin technologies to deliver a spatially cognitive navigation support for visually impaired and neuro-divergent users. Our system captures video input via standard mobile phone cameras, and employs SLAM3R to generate dense 3D point clouds from monocular RGB sequences in real-time. Our custom post-processing algorithm ensures accurate point cloud alignment across multiple viewpoints without requiring predefined reference points. This enhances the capabilities of SpatialLM to produce structured 3D representations, including architectural elements and oriented object bounding boxes. The enriched spatial data is then processed by a locally deployed LLM, which interprets 3D contexts to generate detailed scene descriptions and precise distance measurements between users and surrounding objects. We evaluated our approach across diverse video scenarios featuring various perspectives, looped walking views and captured in multiple environments. The evaluation results demonstrate consistent accuracy in 3D scene interpretation and object localisation, underscoring the potential of our system as a transformative assistive navigation solution that combines advanced visual perception with spatial reasoning

Abstract: Reconstructing 3D scenes under real-world low-light conditions remains challenging due to severe sensor noise, low signal-to-noise ratios, and degraded photometric consistency, which destabilize geometry estimation and novel view synthesis. Existing approaches often rely on well-lit reference data for reliable Structure-from-Motion (SfM) initialization under degraded inputs or apply per-view enhancement methods that introduce cross-view inconsistencies. To address these limitations, we propose \textbf{NOVA-GS}, a unified noise-aware framework for low-light 3D Gaussian Splatting that subsumes enhancement, denoising, and geometry optimization within a single process. Our method leverages VGGT-based feed-forward estimation to obtain robust camera poses and geometry directly from degraded inputs, eliminating the need for SfM. Building on this initialization, NOVA-GS integrates three coupled components: a structure-aware enhancement module for exposure correction, a self-supervised denoising module with blind-spot masking for pseudo-supervision, and a consistency-driven Gaussian Splatting optimization enforcing cross-view geometric coherence. We further introduce a noise-guided spherical harmonic regularization to suppress view-dependent artifacts in noisy regions. Extensive experiments on diverse real-world low-light datasets demonstrate improved geometric fidelity, color consistency, and robustness without requiring paired supervision or well-lit references. https://shaurya2524.github.io/nova-gs/

Abstract: A 3D scene graph groups objects into rooms. When a robot is asked to fetch an object from the kitchen, that grouping is what tells it where to look. An object recorded in the wrong room is not retrievable by a query naming the correct room. We introduce Progressive Boundary Closure, which recovers room layer from a monocular RGB video. A SLAM front end and an open-vocabulary segmenter supply a structural point cloud, camera trajectory and object tracks. The cloud is rasterised into a top-down map, rooms are recovered from it, and each object takes the room holding most of its extent. The difficulty lies in the map itself. Walls are recorded only where the camera looked, so a gap in the boundary may be a doorway or a stretch of wall that was never observed; nothing distinguishes the two. Prior methods treat both as passages, merging rooms that should remain separate. We observe that both require the same treatment: a room should not extend across either, so both are closed and need not be distinguished. Such an opening closes under a small amount of boundary growth, and few sightlines cross it, so points in different rooms rarely see one another. We use the first to recover rooms and the second to assign objects to them. Rooms are obtained by Progressively thickening the boundary inward and freezing each free-space region once it becomes enclosed, so every opening seals at its own scale rather than at a radius fixed in advance. Camera poses are used as seeds, which removes the sampling heuristic and makes the segmentation deterministic. Over 10 floors of 6 HM3D-Semantics scenes, scored against HOV-SG on identical top-down maps, we recover 74 rooms for 72 annotated regions (HOV-SG: 44), raising room F_1 from 0.741 to 0.890 at IoU 0.25 at some cost in precision, and object-to-room ARI from 0.488 to 0.696 (p=0.002, ahead on every floor).

Abstract: Open-vocabulary detection accepts any class list at inference, and promptable segmentation returns regions without class names: the taxonomy has left the model and become an input. Relation prediction has not. Scene-graph models are still trained and evaluated on the 50 or 56 predicates of one annotation style, their relation head conditioned on object labels and so tied to one detector. Three obstacles explain this, none primarily modelling: no relation corpus is both free-text and verified, a label-conditioned architecture cannot accept a vocabulary it was not trained on, and the standard metric rewards agreement with the training corpus, so a larger vocabulary scores as a regression. We present RelateAnything, a 53M-parameter model taking an image and regions from any source and returning scored relations over a predicate vocabulary supplied at inference as strings. Object labels are never an input, so the region source can change without retraining, and the vocabulary is a bank of text embeddings, not a learned classifier. It runs at 20 ms/frame. Training over 19,103 predicates requires positive-unlabeled supervision and a text encoder that separates antonyms, which contrastive encoders embed at cosine 0.95. To supply the supervision we build RA-4M, 474k images and 4.3M relations over 10,102 free-text predicates, generated against numbered box markers and geometrically verified. To measure it we build OV-SGG-Bench, six axes scored across datasets that the priors standard recall rewards cannot satisfy. On three cross-dataset benchmarks and a fourth zero-shot, RelateAnything has 2.3-3.5x the mean recall of the strongest open-vocabulary method of comparable scale, margins that survive a real detector, and leads a 3B-VLM scene-graph model on both metrics at under 2% of its parameters. In-domain measurement overstates transfer gains ~5x. Model, corpus and benchmark are public.

Abstract: Scene-level point cloud understanding remains challenging due to diverse geometries and spatial layouts. While pre-trained 3D point cloud foundation models (PFMs) offer strong transferability, full fine-tuning (FFT) incurs substantial computational and storage costs. Parameter-efficient fine-tuning (PEFT) provides a promising alternative, but existing PEFT methods largely focus on object-level point clouds and overlook serialization-induced partition variations in large-scale scenes. To address this issue, we propose PointPiT, a partition-invariant tuning framework for scene-level point clouds. Specifically, a Scene-aware Structural Adapter (SSA) integrates local geometric patterns with global scene context to mitigate partition-induced representation shifts. Moreover, Gradient Subspace Optimization (GSO) selects informative and partition-stable update directions, suppressing partition-dependent variations during optimization. Extensive experiments across multiple scene-level benchmarks demonstrate that PointPiT achieves competitive or even superior performance to full fine-tuning with less than 1% of backbone's parameters, while achieving consistent state-of-the-art performance among representative PEFT methods.

Abstract: Classical classifier-combination work distinguishes score-level fusion from hard decision-level voting. We revisit this distinction where independently pretrained, frozen foundation models are composed at inference time for generalized few-shot 3D segmentation. We ask: how much useful semantic information is lost when heterogeneous sources are collapsed to a single class before they can interact? We answer with a same-input semantic-retention intervention. Dense RegionPLC and sparse cross-view SAM3 evidence, model weights, masks, geometry, vocabularies, and fusion rules are frozen; only the number of semantic alternatives retained before interaction is varied via a matched top-k ladder. On 156 held-out ScanNet200 scenes, top-1 reaches 28.47 harmonic-mean (HM) IoU while full distribution fusion reaches 34.87 HM (+6.40, 95% CI [+5.24,+7.64]). The pattern replicates on 50 ScanNet++ scenes: 23.02 vs. 26.50 HM (+3.48, 95% CI [+1.64,+5.93]). The conclusion is robust: full-distribution HM is stable across sparse-source weights 0.3--0.7; alternative operators (max, geometric pooling) also outperform top-1; and a GroundingDINO--SAM2.1 source-replacement diagnostic shows monotonic HM increase from 14.77 to 18.75 with full retention. Calibration diagnostics reveal opposite miscalibration of the two sources, yet correcting calibration does not eliminate the retention advantage. Across datasets and source stacks, most information is recovered by retaining a compact set of plausible alternatives. The contribution is a controlled diagnosis of premature semantic collapse as a repeatable information bottleneck in heterogeneous frozen-model composition.

Abstract: Code world models represent worlds as executable programs, but this representation alone does not determine how to construct a complex world. We introduce Recursive Code World Models (RCWM), a framework for reconstructing complex 3D worlds in code from a single reference image. RCWM couples a Recursive Scene Program (RSP) representation with a construction solver that recursively calls itself. An RSP represents the executable world as compositional scene code, while each solver call follows the same complete process: establish the whole, recursively reconstruct unresolved parts, and revisit the whole to refine their composition. This global-local-global recursion gives fine-scale structures their own perception-and-editing loops while preserving scene-wide geometry and relationships. Reference-aligned views propagate a shared camera projection across levels, while parent revisitation addresses boundaries, spatial relations, and shared errors that emerge after local refinement. A vision-language coding agent directly compares reference images with scene renders to guide refinement, recursive descent, and return. Across complex scenes, RCWM outperforms prior code-based image-to-scene reconstruction methods. Ablation studies further support the benefits of recursive construction and suggest that deeper calls can improve finer-scale reconstruction. RCWM provides a recursive construction principle for building complex executable worlds from visual evidence.

Abstract: Optical flow methods typically rely on task-specific inductive biases, such as correlation volumes, feature warping, and iterative refinement, among others, to reach high accuracy. While effective, such biases constrain the model to predefined heuristics, which can limit its expressivity and lead to more complex pipelines and additional computational cost. We present FreeFlow, a hierarchical transformer built without any flow-specific components, using instead a single feed-forward encoder--decoder. FreeFlow combines three attention variants: window attention for local processing, shifted-window attention for cross-window information exchange, and a global attention operating at a reduced resolution. The resulting architecture scales naturally with model capacity, enabling a consistent accuracy gain from small to large variants. Despite the absence of standard inductive biases, FreeFlow achieves state-of-the-art results on major benchmarks, including Sintel (0.68/1.48 EPE on Clean/Final), KITTI-2015 (3.23 Fl-all), and Spring (3.192 1px), while remaining memory efficient at 1080p inference.

Abstract: We introduce Complex-Valued Quadratic Phase Gaussian (CVQPG), a novel hologram representation method that replaces standard 2D Gaussian representations used in 2D Gaussian Splatting with 2D quadratic phase functions. CVQPG incorporates additional learnable parameters to control the curvature of these bases. We evaluate our approach against state-of-the-art methods, exceeding the visual quality by +0.19 dB (RGB) and +0.33 dB (grayscale) on average in holographic reconstructions. Specifically, our equal parameter count evaluations show that modulating the primitive's wavefront is an effective and lightweight enhancement for hologram representations. In addition, our frequency domain analysis illustrates that CVQPG has successfully preserved the mid-to-high frequency band of natural images.

Abstract: Recent visual generators produce high-fidelity images yet often violate physical consistency under ego-motion, limiting their use for spatial reasoning and embodied planning. Existing benchmarks largely focus on isolated images or single-step quality, leaving this challenge underexplored. We introduce EgoGenEval, a geometry-grounded, pose-free benchmark designed to evaluate the physical consistency of visual generators under ego-motion, and organize our study into two parts. (1) EgoGenEval contains 1,400 cases and 2,360 target views spanning single-step and multi-step ego-motion. It separately measures Camera Motion Grounding (CMG) and Scene State Preservation (SSP), with both metrics validated against blinded human judgments. Evaluating 16 pose-free generators together with two pose-conditioned references reveals that current models struggle to execute camera motion while maintaining scene state, and that no system performs well on both axes at once. (2) To examine whether benchmark-derived data can improve these capabilities, we build EgoGen-Train from the same geometry-grounded pipeline and run controlled SFT studies. These show that pairwise supervision does not reliably improve camera-motion grounding and scene-state preservation together: even at the full training pool and the longest budget, scene preservation gains a fraction of what camera motion does. This points to the pairwise teacher-forced objective itself as the binding constraint, motivating a trajectory-centric paradigm that couples self-conditioned rollouts with explicit pose and visibility supervision.

Abstract: Render--match--PnP relocalization establishes correspondences between query image pixels and 3D map points for camera pose recovery, but their potential to support dense depth estimation is often overlooked. To exploit this geometric information, we present RIDE, which estimates dense metric depth from a robot's RGB stream. Given a metrically scaled 3D Gaussian Splatting (3DGS) model, RIDE combines sparse metric depth observations derived from PnP-RANSAC inlier correspondences with the geometric prior of a pretrained video-depth model. To handle uneven and intermittent observations, it integrates global and local depth correction with temporal memory, supporting depth estimation through short observation gaps after metric scale initialization. Trained on public RGB-D videos, RIDE is evaluated on robot sequences without fine tuning. Experiments show improved depth accuracy and temporal consistency over scale-only calibration, demonstrating how localization geometry can support both pose recovery and dense robot perception.

Abstract: Accurate modelling of illumination is central to realistic image synthesis and scene understanding. Yet, there is little exploration into whether image generative models are good at this task or whether physical plausibility remains a key challenge for them. Clearly, significant progress has been made in realistic image synthesis, but do models truly understand lighting in a physically accurate manner? To answer this question, this work proposes a benchmark to assess the lighting understanding and harmonisation capabilities of generative models. Our key insight is that evaluating lighting understanding for such models only requires testing how well they insert novel objects into real photographs whilst maintaining consistent illumination. To do so, we use a multi-illumination dataset with images containing simple objects serving as ``light probes'', and prompt models to inpaint the same object onto the original image, then compare the generated results against the ground-truth light probes. We then estimate the lighting direction, colour and radiance distribution from the inpainted probes, providing a quantitative measure of illumination accuracy and photometric realism. Our work establishes a scalable evaluation protocol to systematically assess how well generative models capture and reproduce real-world lighting, offering a foundation for benchmarking the photometric accuracy of any future models. All code and data are available at https://lvsn.github.io/SheddingLight/ .

Abstract: Monocular 3D hand and body mesh recovery often suffers from severe occlusion and ambiguity. Traditional deterministic methods typically regress a single optimal solution, leading to overconfident predictions. In this paper, we introduce an exploration--exploitation paradigm for ambiguous mesh recovery with multi-hypothesis learning and selection. Specifically, during exploration, based on our probabilistic formulation and entropy maximization, we propose a novel multi-hypothesis method referred to as MHE-Former. It is a Transformer-based multi-hypothesis framework, ensuring high training efficiency and label friendliness while generating plausible and diverse hypotheses. During exploitation, we propose Hypothesis Selection, a context-aware process for multiple predictions. Especially leveraging VLM's powerful visual understanding and reasoning capabilities, it allows users to choose the most plausible and desired estimate with additional evidence and natural language intent. Extensive experiments demonstrate that our framework achieves state-of-the-art performance in accuracy and diversity across multiple datasets. The user preference study further shows the practicality of our hypothesis selection process.

Abstract: Accurate camera intrinsic calibration is fundamental to robot perception, and the accuracy depends on the quality of the collected images. However, existing target-based calibration methods often require the practitioner to manually filter out high-quality images and to specify an appropriate radial distortion order. This paper presents a fully automatic intrinsic calibration pipeline that determines both from the collected data. We adopt an iterative rejection scheme that estimates parameters on a candidate image set and removes views whose mean residual exceeds a multiple of the median. Crucially, this process runs independently under each candidate distortion order, so that the retained image set is consistent with the residual scale of that order. Further, the distortion order is selected on held-out images, with the intrinsics and distortion fixed and only the board pose re-estimated, ensuring that an added coefficient is supported by independent observations. Finally, we integrate both steps into an interactive calibration tool that supports full-pipeline data inspection and parameter estimation. Experiments on our own camera data and five public real-world datasets show that image filtering reduces the held-out reprojection error by 25\%, the order selection further by 5\%, achieving the lowest held-out mean among four compared configurations without manual image selection. We will release the code and data to facilitate future research.

Abstract: Recent advances in Intrinsic Image Decomposition (IID) have increasingly relied on generative models. However, progress remains limited by three key challenges: (a) insufficient physical consistency, (b) high computational cost at inference time, and (c) limited generalization capabilities. In this work, we show that latent bridge matching (LBM) effectively addresses these limitations for albedo estimation. We introduce a novel LBM-based architecture that enforces physical consistency through a pixel reconstruction loss, benefits from the inherent efficiency of LBM low-cost inference, and improves generalization across diverse datasets by incorporating a shading conditioning. In this extended version, we additionally show that conditioning the shading estimator itself on the predicted albedo further improves reconstruction fidelity, and we benchmark our best model against stateof-the-art IID methods across five real and synthetic datasets.

Abstract: Learned image matching has experienced significant progress in recent years, culminating in robust and accurate matchers such as RoMa, whose robustness is often attributed to its use of frozen DINO features. In a parallel development, feed-forward reconstruction models, such as VGGT, have been trained on ever-growing datasets to accurately regress dense 3D point maps and camera poses. The distinction between matchers and feed-forward reconstruction models has become increasingly blurred with the introduction of matching losses in models such as MASt3R and VGGT-$Ω$. This raises a natural question: what do feed-forward 3D models know about image matching? In this work, we answer this question by analyzing three scenarios: (i) zero-shot matching of patch features, (ii) direct matching of 3D point predictions, and (iii) training a full matcher on top of the learned representations. We find that, despite performing poorly in zero-shot matching, especially in later layers, feed-forward reconstruction models provide strong representations for linear probing and full matching pipelines. We further show that, even without any training, their raw predictions alone enable competitive matching, albeit only under moderate viewpoint changes and modality gaps. Based on these insights, we retrain RoMa v2 by replacing its DINO backbone with VGGT-$Ω$. Our resulting model, \ours, outperforms state-of-the-art matchers on a wide range of benchmarks, e.g. +8.1 mAA compared to RoMa v2 on WxBS.

Abstract: Camera pose estimation is a key step in 3D reconstruction and view-synthesis pipelines. We present a deep, global Structure-from-Motion framework based on learned view-graph aggregation. Our method employs a permutation-equivariant, edge-conditioned graph neural network that takes noisy pairwise relative poses as input and outputs globally consistent camera extrinsics. The network is trained without ground-truth supervision, relying solely on a relative-pose consistency objective. This is followed by 3D point triangulation and robust bundle adjustment. Our approach is efficient, scalable to more than a thousand images, and robust to graph density. We evaluate our method on MegaDepth, 1DSfM, Strecha, and BlendedMVS. These experiments demonstrate that our method achieves superior rotation and translation accuracy compared to deep track-centric methods while registering more images across many scenes, and competitive results compared to state-of-the-art classical pipelines, while being much faster.

Abstract: We introduce Point4D, a feed-forward model for 4D reconstruction of long-range video sequences. Point4D is able to reliably infer dense per-point 3D trajectories across multi-hundred-frame videos, unlike existing 4D methods that are limited to short input windows of at most a few dozen frames. A key innovation that enables this is our flexible 3D query-based motion decoder that decouples trajectory prediction from image-plane visibility. The predicted 3D endpoints are then directly re-queried in the next chunk without re-projection or matching. Furthermore, we show that extracting and reusing a visual descriptor from an arbitrary frame where the point is visible leads to better performance than relying solely on the source patch. Overall, Point4D achieves state-of-the-art performance across diverse long-video tracking benchmarks spanning over 200 frames and largely outperforms previous feed-forward 4D method. Project page: https://point-4d.github.io

Abstract: An embodied assistant working beside a person must track task state, recognize help seeking, choose how to intervene, and produce an appropriate response. Existing procedural datasets richly describe individual execution, while interactive datasets capture remote verbal instruction or undifferentiated co-working. They do not jointly link a co-located helper's verbal and physical interventions to performer requests, task state, assistance triggers, and outcomes. We introduce DYAD (DYadic Assistance Dataset), a synchronized multimodal record of human-human assistance during gearbox assembly. Across 20 sessions, one trained helper follows a guidance-first policy while assisting HoloLens 2 wearers. DYAD links 528 task-step intervals and 611 performer requests with 851 valid assistance records spanning verbal and physical help. DYAD's annotations span the assistance process; three reference tasks evaluate selected components rather than an end-to-end system: causal step understanding, pre-onset mode anticipation, and instructor response generation. On 829 eligible mode events, the strongest four-seed RGB mean is 0.548 +/- 0.007 macro-F1; causal metadata reaches 0.624 and a privileged trigger mapping 0.915, revealing information not recovered from pre-onset RGB. DYAD's contribution is not scale, but a linked interaction structure spanning help seeking, intervention choice, execution, and outcome under egocentric and workspace sensing.

Abstract: Language-guided robots need persistent scene memories to follow instructions, revisit objects, and resolve references to objects encountered over time. While much of language-guided scene-memory retrieval has emphasized spatial or relational references, many everyday object references specify objects by multiple persistent attributes, such as category, material, size, or surface appearance. We formalize this problem as attribute-compositional retrieval, where a fixed object-centric scene memory is queried with natural language to retrieve the object satisfying the requested attributes. To investigate this capability directly, we introduce a controlled evaluation protocol with fixed scene memories and attribute-defined targets, separating retrieval from perception and annotation ambiguities. We then propose FRAME, which turns language into query-relevant attribute weights, uses learned readouts to estimate per-attribute evidence from object embeddings, and ranks objects by aggregating this evidence according to the query. Across held-out scenes and object assets, FRAME outperforms representative scene-memory retrieval baselines while reducing post-decomposition object scoring to lightweight matrix-vector computation. These results position attribute-compositional retrieval as a complementary scene-memory capability for language-guided robots, showing that persistent object attributes can be exposed as composable evidence for accurate and efficient multi-attribute retrieval.

Abstract: Reasoning segmentation converts an implicit linguistic conclusion into a precise mask, requiring both semantic identification and spatial grounding. Existing MLLM-segmenter interfaces either use a special trigger or compress both signals into one context, although they receive different supervision and fail differently. This coupling obscures whether a failure arises from target interpretation or from localization. We present SeGDeP, an explicit what-where interface. A semantic prompt branch and an independent geometric projection path transform resolved MLLM states into semantic features and a DETR-predicted box, which jointly condition a SAM 3 mask decoder. Training first aligns this executable interface, then uses group reward-decoupled policy optimization (GDPO) to balance format, box-IoU, and mask-IoU feedback. SeGDeP-4B reaches 82.7 average cIoU over eight RefCOCO-family splits and 66.0/59.6 gIoU on ReasonSeg val/test while adapting only 0.38% of Qwen3-VL parameters through LoRA. Controlled stage-wise ablations, gradient diagnostics, and prompt interventions further show that the two paths develop complementary semantic and geometric specialization rather than duplicating the same evidence.

Abstract: We present FIRE3D, a unified framework that takes a single RGB image or casual RGB video and transforms it into simulation-ready 3D scene assets for games and interactive applications in under a minute. At the core of FIRE3D is a feed-forward, end-to-end network that predicts a compositional scene representation from posed RGB-D observations estimated from the RGB capture, including the 6-DoF pose, bounding box, mesh, and texture for every object. By modeling the scene as a collection of discrete entities, FIRE3D produces amodally complete and simulation-ready environments where objects are physically decoupled and ready for interaction. Our framework requires no test-time optimization, runs orders of magnitude faster than prior interaction-ready methods, and provides object-level completeness beyond existing feed-forward 3D approaches. We demonstrate competitive or state-of-the-art results across pose accuracy, geometry completeness, and texture quality across various datasets while being orders of magnitudes faster. Project page: https://xiahongchi.github.io/Fire3D/

Abstract: Egocentric 4D interaction forecasting aims to anticipate both where future interactions will occur in 3D and how the human body will move to realize them, providing an important capability for assistive robotics and human-computer interaction. Existing methods struggle to translate semantic understanding into precise continuous 3D localization and to balance motion diversity with structural consistency in pose forecasting. More fundamentally, these tasks are often modeled separately, leaving the continuous geometric and temporal correspondence between interaction locations and body motion insufficiently captured. To address these challenges, we introduce Coherent4D, a large-scale egocentric dataset for continuous 4D interaction forecasting, comprising approximately 233K samples across three domains. Each sample pairs a sequence of future 3D interaction locations with corresponding full-body poses, aligned in time and expressed in a shared coordinate system. We also provide evaluation metrics in continuous space. Building on this formulation, we propose HIGFlow, a Hand Interaction Guided Residual Flow framework that models forecasting as a cascaded where-to-how process. HIGFlow first forecasts continuous future interaction locations by combining semantic grounding with short-horizon visual dynamics, and then uses the predicted location sequence to condition a deterministic motion anchor and residual Flow Matching for diverse yet structurally consistent full-body motion forecasting. Extensive experiments across all three domains demonstrate consistent improvements over representative baselines on both location and pose forecasting, while ablations validate the contributions of the proposed components. The project page is available at https://corrineqiu.github.io/from-where-to-how/.

Abstract: A simultaneous localization and mapping (SLAM) method using a monocular camera and a low-cost inertial measurement unit (IMU) sensor is an effective way to fulfill a low-cost sensor configuration. Using this sensor configuration, visual-inertial system (VINS) focuses on fusing data from a camera and an IMU sensor to estimate the six degrees-of-freedom (DOF) of the sensor pose. Typically, VINS uses only a single camera as visual input, which lead to problems such as error accumulation due to occlusion, various illumination, and textureless environments. In this paper, we propose a new multiple fisheye camera-based visual-inertial system called MFVINS. We present an IMU-aided FAST feature tracker for multiple cameras that enables efficient extraction and robust matching of local features. Then, the proposed method filters out outliers caused by fisheye distortion on the normalized image plane. Subsequently, a new reprojection error with physical validity constraints is proposed for bundle adjustment using learning-based depth estimation. The proposed method is applied to various scenarios, and its effectiveness is demonstrated by comparing previous VINS methods. In particular, MFVINS is implemented in real-time process to leverage the advantages of using multiple cameras -- robustness against occlusion and textureless regions -- while reducing the computational burden.

Abstract: We present GOLF, the first-place solution to the SHOW3D Interaction Field Estimation Challenge at HANDS@ECCV 2026. Given synchronized egocentric stereo views, the task is to predict a 3D vector from each of 21 hand joints to the closest point on the manipulated object. GOLF combines dense global context, locally sampled hand/object evidence, and common-frame Plücker-ray geometry. We adapt DINOv3 ViT-H+/16 with LoRA and trainable LayerNorm parameters, then jointly decode both interaction fields. Our primary model achieves an official score of 27.61 and a mean ADE of 27.96 mm on the hidden test set. An equal-weight ensemble with a complementary directly fine-tuned variant improves these results to an official score of 27.47 and a mean ADE of 27.82 mm, securing first place.

Abstract: Gaussian splats provide a fast, high-fidelity representation for 3D objects but are often constructed from incomplete input data in practice, leaving missing regions. Existing completion methods either do not preserve the original splats or require scarcely available 3D training data. We propose GSComplete, which combines 3D generation based on Score Distillation Sampling with a novel preservation loss that encourages the original splats to be preserved where they should be visible. This effectively completes the Gaussian splat object using only 2D diffusion priors while fully preserving existing splats and generating new splats only in missing regions, without occluding the input. To evaluate our approach, we introduce a new dataset of partial Gaussian splat objects and show that GSComplete achieves significantly more accurate preservation of the input than existing methods with comparable plausibility of the completed result. Our code and dataset will be made available upon acceptance.

Abstract: We present EdMCGS (Event-driven Markov chain Gaussian Splatting), an end-to-end method for reconstructing dynamic 3D scenes from extreme-low-frame-rate RGB together with an event stream, which can then be rendered at any intermediate timestamp. Methods relying solely on RGB images generate numerous artifacts due to the lack of evidence from between consecutive frames. To supply this missing evidence, we model the scene motion as an event-driven Markov chain, in which the sparse RGB frames anchor the state at their own timestamps while the events recorded within an interval drive the transition across it. Since the transition reads the events of the current interval, it remains active at inference and produces the in-between motion of the 3D Gaussians directly from the events rather than by interpolation, which sets our method apart from prior work that uses events only as training-time supervision. The state is carried by a compact set of control points, each driven by the events sampled in the neighborhood of its own image projection, and a temporal local isometry term keeps the propagated motion locally rigid. Experiments on synthetic and real-world scenes show that EdMCGS outperforms both RGB-based and event-based baselines, while rendering in real time with far fewer Gaussians than the strongest event-based baseline. We release our source code and a new dataset at https://github.com/joseclipse/EdMCGS.

Abstract: SLAM systems based on 3D Gaussian Splatting (3DGS) have recently demonstrated promising reconstruction accuracy for dense 3D scene representations. However, current 3DGS systems struggle to meet the strict demands of real-world deployments due to severe limitations in operational performance and map adaptability. To this end, we propose LightSplat, a hybrid-representation RGB-D SLAM framework. It synergizes local sparse features for robust and fast tracking with a dual-thread backend that progressively constructs dense Gaussian submaps. Crucially, we enable online loop closure through feature-accelerated 3DGS registration, refining overall map consistency through pose graph optimization. Ultimately, LightSplat achieves the online reconstruction of high-fidelity Gaussian map. Extensive experiments on multiple datasets and real-world robotic platform demonstrate that our method achieves near state-of-the-art reconstruction quality and the capability to accommodate practical camera motions, maintaining an average framerate of 8 FPS. Overall, LightSplat provides an efficient and robust foundation for deploying high-fidelity 3DGS in real-world environments.

Abstract: Despite the rapid progress of Multimodal Large Language Models (MLLMs) in 2D vision-language tasks, robust multi-view spatial reasoning remains a fundamental bottleneck due to the lack of structured 3D cognitive pathways in existing datasets. To address this, we introduce MV-STRIDE, a Multi-View hierarchical SpaTial Reasoning dataset with Interdependent and DEcomposed capabilitiEs. Moving beyond flat data structures, MV-STRIDE explicitly models the dependency relationships between foundational perception, scene understanding, and complex contextual reasoning, providing a coherent learning pathway aligned with human spatial cognition. We develop a systematic QA generation pipeline leveraging diverse 3D scene sources that enforces cross-view dependency constraints to prevent single-view solvability, generating multi-level spatial reasoning tasks supported by cognitively grounded chain-of-thought supervision for complex inference. Extensive evaluations demonstrate that our multi-stage training framework based on our hierarchical dataset achieves state-of-the-art performance across multiple spatial reasoning benchmarks, notably the multi-view oriented MMSI-Bench. Our approach enables MLLMs to maintain robust, 3D-consistent spatial reasoning across diverse viewpoints. The code and dataset are available at https://co1dspring.github.io/MV-STRIDE/.

Abstract: Text spotting requires both accurate text recognition and precise spatial localization. Current specialised spotters excel at predicting tight bounding boxes in natural scenes, but falter on complex or artistic text, whereas multimodal large language models (MLLMs) possess strong recognition capabilities yet remain weak at localisation. To equip the text spotter with general and powerful recognition capabilities and to maximize its localization ability, we explore two MLLM-based fine-tuning methods: Supervised Fine-Tuning (SFT) and reinforcement learning fine-tuning based on Group Relative Policy Optimisation (GRPO). An interesting finding is that SFT is less effective than GRPO at enhancing recognition, while GRPO is less effective than SFT at enhancing detection. To compensate for each other's shortcomings, we introduce a joint training strategy, SupGRPO, which simultaneously optimizes the model using both SFT and GRPO. SupGRPO employs the specially designed reward functions and develops a matching-based online SFT applied solely to coordinate tokens. It both mitigates the reward sparsity problem of GRPO and avoids the instance order dependency problem of SFT. To evaluate particularly challenging cases, we curate ATS, a dataset for artistic text spotting. Experiments demonstrate that SupGRPO improves both text recognition and detection, and attains superior performance. Our code and dataset will be released at https://github.com/Psycho-9/SupGRPO.

Abstract: Predicting hand--object interaction fields requires locating the nearest object-surface point for each hand joint, often from small and partially occluded image regions. We view this task as joint-conditioned surface-endpoint estimation: each joint has its own nearest endpoint, while endpoints from the same hand can draw on shared local surface evidence. This structure motivates Joint-Conditioned Stereo Surface Reasoning (JSSR). A temporal-stereo network jointly predicts 3D joints, a direct interaction field, and per-view endpoint evidence. Calibrated candidate search evaluates endpoint hypotheses using joint-specific image compatibility and cross-view correspondence. A hand-shared candidate support lets joints draw on common surface evidence, and a learned residual gate controls the geometric correction when observations are ambiguous. Our system built on this method ranked third on the SHOW3D Interaction Field Challenge leaderboard.

Abstract: Reasoning over language instructions in embodied tasks such as robotics often requires understanding spatial relations from a speaker's situated perspective. Humans infer such perspectives from shared environmental knowledge, activity context, and commonsense. Recent vision-language models (VLMs) appear capable of spatial reasoning, but their ability to infer a speaker's viewpoint from contextual cues and interpret situated spatial relations from that viewpoint remains unclear. We call this capability contextual observer grounding. To study this capability, we construct the Point-of-View Benchmark (POVBench), a dataset of 3D scenes and queries that disentangles Inferred, Stated, and Given forms of observer grounding in natural embodied communication. Given multi-view observations and a natural-language sentence, models must localize unseen or underspecified targets from situated spatial and contextual cues. Across state-of-the-art VLMs, localizing targets from directional language remains challenging, even when observer grounding is made explicit. We find that explicit breakdowns of observer-relative spatial reasoning improve target localization. Our project page is available at https://mimo-owl.github.io/POVBench/.