3D VLMs

Large Multimodal Models (LMMs) have achieved remarkable success on images and short videos, yet scaling them to long videos remains challenging due to frame-centric tokenization and limited context windows. 3D geometry provides a natural compression mechanism for visual streams: depth and camera pose enable observations from multiple views and time steps to be fused into a persistent, world-aligned representation. While recent 3D LMMs leverage geometry-aware representations to improve spatial reasoning, they continue to lag behind specialist 3D perception systems on grounding and segmentation tasks. We argue that a key limitation is geometry-aware decoding: existing methods communicate 3D predictions through language tokens, proposal selection, or lightweight grounding queries, creating a bottleneck between language reasoning and dense geometric prediction. Building on these insights, we introduce Qwen-3D, a geometry-aware LMM that compresses visual information within the Qwen backbone using multi-view geometric cues, enabling efficient long-horizon visual reasoning over static scenes. Qwen-3D augments visual tokens with 3D Rotary Positional Embeddings, allowing attention to operate directly in 3D scene space rather than across independent image frames and thereby facilitating scalable cross-view and temporal reasoning. To bridge language and geometry, Qwen-3D incorporates a query-based segmentation decoder that grounds language directly in the underlying 3D scene representation, unifying referential grounding, instance segmentation, and visual question answering across both images and videos. Across a diverse set of benchmarks, Qwen-3D surpasses existing 3D LMMs and outperforms several large proprietary 2D models. Notably, Qwen-3D achieves these improvements while maintaining strong performance on standard 2D vision-language benchmarks by jointly training on 2D and 3D data.

Recent advances in image generation have demonstrated the potential of unified multimodal models that integrate understanding, generation, and editing. However, unified 3D modeling remains constrained by scarce multimodal data, particularly the lack of large-scale and geometrically consistent editing data. To address this limitation, we propose Hunyuan3D-Buffalo 1.0, a unified framework supporting 3D understanding, text-to-3D generation, instruction-guided 3D editing, and text-grounded part generation within a single architecture. To enable scalable training, we construct an 87M-scale 3D multimodal corpus, comprising 25M understanding samples, 50M text-to-3D pairs, and 12M editing pairs generated using Nano3D-v2. Architecturally, the framework combines Hunyuan3D-VLM for semantic, structural, and spatial understanding with Hunyuan3D DiT for high-fidelity 3D synthesis. The VLM provides multimodal semantic conditions for generation, while editing and part generation additionally condition the diffusion process on the source object representation to preserve its overall structure and unedited regions. Extensive experiments show that Hunyuan3D-Buffalo 1.0 achieves state-of-the-art or leading performance on text-to-3D generation and 3D editing benchmarks, while exhibiting strong understanding and part-generation capabilities. Our analysis further shows that both generation and understanding improve editing, demonstrating the effectiveness of unified 3D multimodal training. Project Page: https://tencent-hunyuan.github.io/Hunyuan3D-Buffalo1.0/

Recent 3D vision-language models (3D VLMs) construct geometry aware tokens by projecting 2D visual features into world coordinates, enabling spatial reasoning for tasks such as 3D question answering. However, this design generates thousands of tokens per scene, resulting in substantial computational and memory overhead. While token compression has been extensively studied in 2D VLMs, existing approaches rely on semantic relevance or attention-based selection that overlook the structured spatial nature of 3D tokens. Moreover, redundancy in 3D representations cannot be resolved by spatial proximity alone, as object-level token imbalance persists even after spatial aggregation. To address this, we propose 3DZip, a three-stage token compression framework that first applies coarse voxelization to remove point-level redundancy, then selects anchor tokens based on feature-space diversity via a Determinantal Point Process, and finally merges remaining tokens under spatial constraints to preserve geometric coherence. Experiments on three 3D question answering benchmarks demonstrate that 3DZip consistently outperforms existing compression methods, retaining 94.7% of the original performance with only 128 tokens, achieving a $1.92\times$ faster inference speed.

3D scene understanding requires reasoning about entity existence, spatial layout, and object relations, yet RGB images alone often provide insufficient 3D cues. Existing 3D-VLMs commonly rely on depth or 3D-position-aware inputs at inference time, introducing additional acquisition, reconstruction, or annotation costs that limit RGB-only deployment. We therefore study how training-time 3D evidence can be converted into spatial reasoning capabilities retained under RGB-only inference. We propose a privileged-evidence distillation framework that constructs a distillable teacher through a unified evidence interface and controlled residual injection, and transfers its knowledge to a deployable student receiving only RGB images and questions through logit and structured representation distillation. To avoid imitating teacher signals unsupported by RGB, we further introduce evidence-sensitivity-guided distillation, which uses corrupted evidence to identify highly evidence-dependent targets and down-weight their supervision. We also define a recoverability decomposition based on the matched baseline, teacher, and student, separating privileged gains into RGB-recoverable improvements and residual teacher advantages. Across four benchmarks, the teacher achieves the best result on 7 of 11 reported metrics among the compared methods. The RGB-only student outperforms its matched baseline on all 11 metrics, including gains of 10.4 ScanQA CIDEr and 19.1 Scan2Cap [email protected], without additional inference-time inputs. These results validate the effectiveness of training-time privileged 3D evidence distillation for both teacher performance and deployable RGB-only spatial reasoning. Separately, our matched baseline-teacher-student analysis characterizes privileged-gain transfer across evidence types and spatial skills.

3D point cloud-language models (3D-LLMs) enable 3D understanding by pairing point cloud encoders with large language models, but existing methods rely on costly multi-modal encoders (e.g., ULIP-2) that require image-text-point cloud alignment on 8x A100-scale compute, creating high barriers for research and deployment. In this work, we systematically investigate whether low-cost self-supervised point cloud encoders, specifically PCP-MAE and Point-MAE, can serve as effective alternatives. Using MiniGPT-3D as our testbed, we evaluate 7 encoder initialization/pre-training setups (1 multi-modal baseline, 5 self-supervised, 1 random init) under frozen and unfrozen fine-tuning (12 total groups), across 2 architectures (MaskTransformer, PointTransformer), 3 objectives (PCP-MAE, Point-MAE, random init), and 2 datasets (Objaverse 660K, ShapeNet55-34 approximately 50K). Our experiments reveal three key findings: (1) The four-stage MiniGPT-3D pipeline can effectively train a 3D encoder from random initialization: an end-to-end trained random init encoder reaches 52.50% open-vocabulary accuracy and 44.45 captioning score, approaching top pre-trained variants; (2) Architecture and pre-training objective show strong crossover interaction: PCP-MAE + MaskTransformer achieves 59.00% accuracy (best self-supervised), while Point-MAE + MaskTransformer drops to 46.50%, with the pattern reversed for PointTransformer; (3) Closed-set ModelNet40 classification remains a core weakness of purely geometric encoders, reaching only ~13-18% accuracy vs. ~62% for the multi-modal baseline, even after end-to-end fine-tuning. Our results offer practical guidelines for cost-effective 3D-LLM design and reveal interaction patterns between self-supervised objectives and encoder architectures.

Integrating 3D medical images with vision-language models (VLMs) holds substantial promise for computer-aided diagnosis. However, volumetric images generate prohibitively long visual-token sequences with considerable spatial and inter-slice redundancy. Existing token compression methods typically apply uniform reduction or rely on a single importance signal, increasing the risk of removing regions that are clinically relevant to the query or structurally distinctive. To address this limitation, we propose MedARC, a unified, training-free framework for Adaptive Redundancy Compression of visual tokens in 3D medical VLMs. MedARC estimates token importance by integrating three complementary cues: self-attention from the VLM vision encoder, which reflects the model's intrinsic visual focus; similarity between projected visual tokens and text embeddings, which identifies query-relevant regions; and deviations of local visual foundation model features from the volume-level feature center, which highlight structurally distinctive anatomy. The resulting importance distribution guides a saliency-aware merging strategy that preserves informative tokens while consolidating redundant ones rather than simply discarding them. Experiments on CT-RATE and MR-RATE show that MedARC reduces visual-token overhead and inference time while preserving or improving diagnostic performance. Its multi-cue scoring cost is outweighed by the savings from processing fewer tokens, with greater benefits expected for larger language models.

Multimodal large language models (MLLMs) hold immense potential to revolutionize clinical practice, yet deploying them in the medical domain is fundamentally a vision-centric challenge: models must absorb knowledge from heterogeneous 2D and 3D medical images, and evaluation protocols must align with radiologists' clinical practice and provide an accurate, fine-grained and factualness-driven assessment. In this paper, we introduce ClinFusion, a vision-centric MLLM designed for holistic medical understanding that systematically addresses these limitations. We propose a compositional and cascaded vision encoder architecture featuring a Cascade Spatial-Aware Locality Fusion operator that unifies diverse 2D and native 3D medical image understanding within a fused encoder. We further introduce a vision-grounded evaluation framework, including MedIF-Bench for instruction-following assessment and a region-of-interest-grounded method for clinically aligned and factualness-driven report generation evaluation. We show that ClinFusion sets a new state-of-the-art across a comprehensive suite of 2D and 3D multimodal medical benchmarks---spanning visual question answering, report generation, and instruction following---as well as textual medical tasks, outperforming leading open-source medical MLLMs (\textit{e.g.}, Hulu-Med, Lingshu) on 20 out of 24 benchmarks and demonstrating multimodal capabilities better than powerful proprietary models such as GPT-5.2 and Gemini-3-Flash on 13 out of 16 benchmarks, and can be further augmented with agentic tool use for retrieval-augmented and tool-assisted clinical workflows. A blinded evaluation by board-certified radiologists confirms that ClinFusion produces the highest-ranked reports, and validates our RoI-grounded metric as achieving the strongest correlation with expert judgment among all automatic evaluation metrics examined.

Medical multimodal large language models (MLLMs) are increasingly expected to perform complex image understanding tasks, yet their reliability is often compromised by frequent errors in visual interpretation. To systematically trace these failures, we traverse the hierarchy from high-level clinical tasks down to fundamental visual perception. We therefore introduce Perception-Bench, a large-scale benchmark comprising 1.13 million samples that assesses medical MLLMs across six dimensions: attribute judgment, spatial grounding, spatial understanding, disease prediction, anomaly detection, and report generation, spanning both 2D and 3D radiology images. Our analysis on Perception-Bench reveals that existing MLLMs lack the ability to capture even the most basic lesion attributes, such as location, size, and density. This inability to ground clinical outputs in primary visual evidence reveals that the models' diagnostic unreliability is rooted in a critical but overlooked bottleneck in low-level visual perception. Motivated by this, we propose RadSight, a perception-driven MLLM built upon a dual 2D/3D encoder architecture that preserves native imaging spatial structures. RadSight formulates medical image understanding as a four-stage progressive process: visual-language alignment, fine-grained visual perception, clinical diagnosis, and diagnostic interpretation. The model is trained on an 8.37 million perception-oriented corpus using progressive curriculum learning. On Perception-Bench, RadSight consistently outperforms existing MLLMs across all six evaluation dimensions, with particularly strong gains in spatial grounding and clinical diagnosis. It also achieves consistent improvements on public 2D and 3D medical benchmarks, further demonstrating that robust low-level visual perception is a critical foundation for reliable clinical understanding. Code and model will be publicly available.

Picking the frozen image encoder for a 3D~CT vision--language model (VLM), together with the token-compression scheme on top of it, is a search over many candidates. There are several encoders, several ways to compress their tokens, and several token budgets, and the combinations grow fast. Comparing them the usual way means fine-tuning a large language model (LLM) on each combination, and running the whole sweep this way needs far more compute than most groups can spend. We ask whether a cheap probe on the encoder's cached embeddings can stand in for that comparison. We build an image-grounded probing benchmark over (encoder $\times$ compression) cells, with clinical attribute families and two validation gates, scale-sanity and probe-separability, that keep each attribute well-scaled and decodable. These gates are the main methodological contribution. On this benchmark we compare a range of read-out heads, and in a preliminary study we pair each probe with its matched downstream task. The early signal is encouraging: the cheap probe orders the candidates in close agreement with expensive fine-tuning, at about $r\approx0.95$ on the cells measured so far. We read this as an ordinal claim, a ranking predictor rather than an exact estimate, and we are explicit about where it stays preliminary. If it holds up, encoder and compression choices can be screened in minutes with frozen-token probes, with full training spent only on the finalists.

Despite rapid progress, most existing vision-language models (VLMs) built from 2D visual inputs often struggle when handling various 3D tasks that require fine-grained spatial understanding and reasoning. To bridge this gap, we present VLM-IE3D, a unified framework that enhances the 3D spatial awareness of VLMs by equipping them with both implicit and explicit 3D geometries learned from RGB videos. Our VLM-IE3D introduces Implicit Geometry Tokens (IGTs) that capture high-level geometric priors from input videos, as well as complementary Explicit Geometry Tokens (EGTs) that encode detailed geometric structures from reconstructed 3D attributes. On top of that, VLM-IE3D comes with a 3D-aware adapter that effectively fuses the two types of geometric representations with 2D visual cues. This RGB-only design injects strong 3D inductive biases for fine-grained spatial understanding and reasoning without requiring any additional 3D inputs. Extensive experiments show that VLM-IE3D achieves superior performance consistently across various 3D tasks including 3D video detection, 3D visual grounding, 3D dense captioning, and spatial reasoning. Code and models are available at https://github.com/Vegetebird/VLM-IE3D.

Recent advances in Multimodal Large Language Models (MLLMs) have triggered the development of end-to-end MLLMs for autonomous driving. However, the main emphasis to date has been for MLLMs using 2D images and videos. In contrast, this paper considers MLLM effectiveness using 3D sensors, particularly LiDAR and stereo cameras. LiDAR presents unique challenges to integration within an MLLM, largely because of data sparsity and lack of a grid structure for the data. For similar reasons, fusion of camera and LiDAR data within an MLLM pipeline is also uncommon. However, most autonomous systems rely on LiDAR-based sensing, and incorporating 3D data has been proven to improve performance in traditional 3D scene perception tasks. This paper presents D3VL, a novel MLLM framework that integrates 2D and 3D time-series data in a single but simple architecture. The model aims to answer questions involving traffic scene understanding and safety. D3VL shows an 11% improvement in the KITTI Question-Answering (QA) dataset compared to baseline methods in processing 2D and 3D time-series data. This paper further introduces the Waymo QA dataset extension, which assesses models' capabilities in processing 3D and time-series data under diverse driving conditions. D3VL implementation code and WaymoQA extension can be found on our supplemental website: https://automotivesafety-lvlm.github.io

3D CT vision-language models (VLMs) classify abnormalities from text prompts in a zero-shot manner, enabling cross-institution deployment where labels are scarce and clinical tasks shift faster than supervised models can be retrained. A real CT scan, however, typically contains several co-occurring abnormalities, and the reliability of zero-shot multi-label prediction under distribution shift remains poorly understood. Test-time adaptation (TTA) updates a model on unlabeled target scans without source data or target annotations, yet existing TTA methods target multi-class softmax prediction on natural images or 2D medical segmentation, and none addresses unsupervised multi-label adaptation for zero-shot 3D CT VLMs. We study when TTA helps zero-shot 3D CT VLMs. A controlled diagnostic analysis shows that TTA is conditional: the volumetric input must preserve the encoder's depth structure, and the base representation must transfer to the target cohort, with depth reduction alone lowering internal AUROC by more than 0.12. We then focus on the regime where the base model already separates present from absent abnormalities. We introduce CARVE (Cardinality-Aware Retained-View Entropy), the first TTA method for this setting. CARVE estimates a sample-specific positive-label cardinality $\hat{k}$, optimizes a top-$\hat{k}$ objective to preserve co-occurring abnormalities, and performs memory-efficient multi-view adaptation by scoring weak 3D views without gradients before updating on a retained subset. Across contrastive CT-CLIP and anatomy-aware fVLM, CARVE provides the most consistent improvements across multi-label, three-class, and binary CT tasks when the base model is already discriminative. These results establish multi-label TTA for zero-shot 3D CT VLMs as a distinct problem and CARVE as a cardinality-aware solution.

Recent advances in large language models (LLMs) and their extension to vision-language models (VLMs) have made it easier to combine text and images for tasks such as report generation. Existing VLMs in medicine typically focus on 2D images (chest X-rays), and their extension to 3D imaging has been difficult because of the lack of paired 3D imaging-text data. Thus, we introduce a new method for creating a 3D image-text dataset for brain oncology using 3D MRI scans of glioma and meningioma cases. We use a cooperative system in which several LLMs work together to generate and check reports, ensuring that they are accurate and clear. By leveraging the new 3D MRI-text dataset, we further build a VLM that converts MRI scans into tokens and aligns them with text instructions. Our VLM performed better in report generation and visual question answering tasks than other 2D and 3D methods. Our method not only improves the quality of reports but also helps with better diagnosis and treatment in brain oncology.

Although multimodal large language models (MLLMs) have achieved remarkable progress, understanding 3D spatial relationships from 2D images remains a critical challenge. Existing methods primarily rely on symbolic text tokens, which inherently lack the fidelity to represent continuous geometric information. While recent methods use latent representations to enhance reasoning, relying on a single latent type cannot adapt to the diversity of spatial tasks, leading to misalignment in complex geometric scenarios. To address these limitations, we propose GeoAnchor, an interleaved text-latent reasoning framework. GeoAnchor decomposes 3D spatial information into three complementary components: position latents for object grounding, direction latents for relational orientation, and geometry latents for scene structure. These components are recombined in a structured space to construct local evidence while capturing global context, enabling dynamic and interpretable reasoning. Furthermore, we introduce a collaborative training strategy that guides the model from local spatial perception to comprehensive 3D understanding. Extensive experiments on diverse and complex 3D reasoning tasks demonstrate that GeoAnchor outperforms the state of the art, validating its effectiveness and generalization capabilities.

Open-vocabulary 3D scene understanding is commonly achieved by embedding 2D vision-language features such as CLIP into a 3D Gaussian Splatting scene, turning it into a text-queryable semantic field. However, attaching a high-dimensional feature to each of millions of Gaussians inflates a single scene to gigabytes, which makes storage and deployment the real bottleneck of these fields. Existing compact methods each learn and ship a per-scene codec, an autoencoder, a quantized codebook, or a distilled feature field, entangling field construction with field storage and never compressing the per-Gaussian assignment that holds the bulk of the cost. We argue that construction and storage should be decoupled, and that storage is a rate-distortion problem over the per-Gaussian binding to a small anchor table, a structure no prior open-vocabulary method compresses. We present CoSAG, which constructs the field without any per-scene training through a closed-form transmittance-weighted lift, spatially grounded semantic anchors, and multi-view denoising, and stores it with a spatially predictive entropy coder that ships no decoder. Because the anchors are spatially grounded, the binding is predictable and therefore highly compressible. The transmittance-weighted lift and multi-view denoising yield a clean, view-consistent assignment, so the entropy coder spends almost no rate on correcting noise and instead codes only the residual against its spatial prediction. CoSAG reaches sub-megabyte storage while matching or exceeding the state of the art across the 2D-rendered, 3D-selection, and dense-LSeg protocols, reducing field size by 37 to 76x relative to LangSplatV2 at higher accuracy.

Unified 3D foundation models aspire to generate 3D assets and reason about them in language within a single backbone, but their text-3D interaction remains largely implicit. Existing methods concatenate text and 3D tokens into a flat sequence and rely on self-attention, collapsing coarse structural cues and fine geometric details into one undifferentiated representation. We introduce ELSA3D, a unified 3D model that addresses this with elastic semantic anchoring, structuring language and geometric reasoning jointly along matched abstraction scales. ELSA3D represents geometry with a scale-aware octree tokenizer and introduces Anchor Tokens, sparse cross-modal units that select semantic cues, route them to the most relevant 3D scale, retrieve scale-specific geometric evidence, and write the fused signal back into the unified representation, keeping interaction sparse yet precise. A lightweight per-block router makes both computation and reasoning elastic, choosing which text tokens instantiate anchors at which geometric scale so that cross-modal capacity concentrates where alignment is most needed. ELSA3D achieves state-of-the-art performance across image-to-3D generation, text-to-3D generation, and 3D captioning, outperforming the strongest unified baseline while roughly halving FLOPs and inference latency relative to the non-elastic version of the same model.

Existing 3D scene-grounded Large Language Models (3D-LLMs) focus on answering questions grounded in simplified single-room 3D scenes, lacking the ability to reason over real-world household environments containing multiple interconnected rooms and diverse object categories. We introduce CAIRN, a topology-aware 3D-LLM for multi-room 3D scene understanding. CAIRN aligns transformer attention with scene hierarchy, giving the model explicit awareness of object-level relations and room-level connectivity. It enriches object tokens with room-local relational context via a graph neural network, introduces learned room tokens for room-level abstraction, and applies a hierarchical attention mask with geometric bias to route information according to scene topology. CAIRN is developed on CAIRN-MR, a benchmark we introduce on HM3D for multi-room 3D scene understanding, covering grounding, captioning, and four question-answering tasks that progressively evaluate from intra-room perception to cross-room reasoning. Experiments show that CAIRN outperforms prior 3D-LLMs by a large margin across all CAIRN-MR tasks while remaining competitive on five single-room benchmarks.

Vision-language models (VLMs) have made interactive digital museums increasingly feasible by connecting 3D digitization with natural-language artifact exploration. However, in cultural heritage domains such as ancient Greek pottery, reliable VLM assistance is limited by two challenges. First, open-ended interpretation requires grounding fine-grained 2D/3D visual evidence in specialized curatorial knowledge, yet the retrieval process may introduce weak sources and unverifiable references. Second, when the available evidence is incomplete, noisy, or ambiguous, VLMs often produce confident but unsupported answers instead of calibrated uncertainty. To address these challenges, we propose VaseMuseum, a lightweight and modular multimodal agent framework for intelligent digital museums of ancient Greek pottery. VaseMuseum combines an interactive virtual museum with VaseAgent, which supports both 2D images and 3D artifacts through multimodal perception, 3D-aware reasoning, external knowledge retrieval, and inference-time reliability control. Specifically, VaseAgent retrieves evidence from authoritative web and museum knowledge sources, and source-level control selects diverse and verifiable evidence before generation. Meanwhile, response-level control checks generated claims against the evidence pool and encourages neutral, evidence-bounded answers when support is insufficient or conflicting. Moreover, a training-free GRPO-style selection mechanism favors responses with valid references and calibrated confidence without updating the VLM backbone. Experiments in a realistic digital museum simulation show that VaseMuseum improves citation validity, reduces hallucinations on knowledge-intensive queries, and produces more neutral answers under ambiguity compared with search-enabled VLM baselines.

3D dense captioning, an emerging vision-language task, aims to generate descriptive sentences for each object in the 3D scene. Despite the impressive results achieved by previous methods, they suffer from two limitations. First, current research often employs global rigid transformations, such as rotation, to augment scenes without changing their spatial layouts. However, diverse spatial layouts are crucial for training a 3D dense captioning model to describe spatial relations between objects. Second, previous works mainly focus on the design of the caption generation pipeline while utilizing a simple network architecture for other components, i.e., backbone and detection head, which is crucial for extracting rich semantic information for captioning. In this paper, we propose PVCap to alleviate the aforementioned problems. Our PVCap consists of PseudoCap and VoxelCapNet. Specifically, PseudoCap employs a random mixing technique on instances within the dataset, generating numerous pseudo frames with diverse spatial layouts at the instance level. By utilizing a teacher-student framework, PseudoCap obtains pseudo caption labels for these pseudo frames. This data augmentation approach significantly increases the number of training samples and enhances the model's ability to describe the environment effectively. Regarding VoxelCapNet, we introduce a robust caption network that utilizes voxel features and adapts the caption head to the voxel-based network architecture. Our VoxelCapNet can serve as a competitive baseline for future research on 3D dense captioning. Extensive experiments are conducted on two prevalent benchmarks, i.e., ScanRefer and Nr3D. Notably, our method surpasses current state-of-the-art by 11.41% and 13.99% in [email protected], respectively. Codes will be made publicly available.

Recent Multimodal Large Language Models (MLLMs) struggle to bridge the representational gap between 2D semantic understanding and 3D spatial geometry. Existing 3D-aware models either rely on costly 3D-specific data or utilize RGB-only inputs with heuristic sampling and monolithic, shallow fusion, which respectively disrupt essential spatiotemporal connectivity and induce modality contention across diverse spatial tasks. To overcome these bottlenecks, we introduce SpaR3D-MoE, an end-to-end framework that enables adaptive spatial reasoning by equipping MLLMs with geometry-aware capabilities from only sparse RGB inputs. First, we propose an adaptive spatiotemporal manifold sampling mechanism that constructs a geometry-aware spatiotemporal graph to extract informative keyframes, effectively mitigating sequence redundancy while preserving the scene's topological connectivity. Second, we introduce the heterogeneous geometry-inductive Mixture-of-Experts driven by an instruction-pose aware router, which adaptively routes multimodal tokens to specialized experts, resolving the cross-modal contention inherent in monolithic fusion. Extensive experiments on VSI-Bench, ScanQA, and SQA3D demonstrate that our method achieves state-of-the-art performance. Notably, SpaR3D-MoE achieves the highest average score of 63.5 on VSI-Bench, outperforming the strongest baseline by 7.8 absolute points, alongside relative improvements of 35.4% and 51.4% in Route Plan and Relative Direction tasks, respectively.

Natural-language queries about 3D environments become actionable when responses are verifiable and metric. Verifiability requires explicit grounding to the referred 3D region, while metric answers report physical measurements in real-world units (e.g., size, thickness, clearance, and distance). Existing 3D large multimodal models (LMMs) approaches remain limited: conversational systems typically respond without explicit 3D grounding, while 3D grounding models are not designed for interactive, metric-aware dialogue. In this paper, we present Ground3D-LMM, a unified model that takes a point cloud and an optional RGB image as input and supports 3D spatial conversation with (i) point-grounded responses and (ii) metric numeric outputs at both object and part granularity, including multi-object queries. To evaluate this intersection of grounding and measurement, we define the 3D Grounded Measurement task, which requires predicting the referred 3D region and the corresponding metric quantities in real-world units. We introduce a large-scale dataset built on ScanNet and ScanNet++ datasets with dense object and part annotations and roughly 2.5M question-answer pairs spanning eight tasks, along with a manually verified test set. Extensive experiments on multiple datasets and tasks show that our proposed Ground3D-LMM model provides a strong baseline for grounded, metric-aware 3D conversational understanding. Our dataset and model are publicly available.

Human spatial understanding arises from jointly perceiving geometry and semantics, enabling consistent object identification and localization across viewpoints and time. Current video segmentation models depend on explicit object appearance memory banks for instance tracking, yet they remain vulnerable to large viewpoint changes and long-term occlusions. Leveraging the spatial consistency afforded by modern feed-forward 3D reconstruction models, we propose the Geometry Grounded Tracking Anything Model (G$^2$TAM), a unified framework for promptable instance tracking in 3D using only unordered RGB images or videos. G$^2$TAM employs spatially aligned geometric representations as implicit memory, ensuring stable instance identity and localization across frames and views. At its core is a cross-modal spatial encoder that integrates visual and textual prompts into a shared geometric space, enabling end-to-end spatial reconstruction and instance-consistent mask prediction. To support training and evaluation, we construct InsTrack, a large-scale dataset with a dedicated validation split for benchmarking. Extensive experiments show that G$^2$TAM delivers strong cross-view consistency, promptable instance spatial tracking, video object segmentation and spatial reconstruction, establishing a foundation for interactive, geometry-grounded spatial reasoning.

Point cloud-based understanding has become an important enabler for facility operation and maintenance involving indoor building components. However, existing methods output only discrete labels without explaining component functions or natural language interactions. This paper proposes Building-MLLM, a point cloud-centered multimodal large language model (MLLM) for indoor components, which models point clouds and instructions to generate responses across Simple Recognition, Complex Captioning, and Multi-Engineering Question Answering tasks. Building-MLLM addresses semantic concentration through four domain-specific mechanisms: Point Information Enhancer for task-relevant semantics, Geometry-Preserving Regularization preventing geometric erosion, fixed textual prefix for domain stabilization, and multi-dimensional LoRA balancing recognition with reasoning. A multi-constraint progressive instruction-generation engine is developed to compile a synthetic point cloud-text dataset with 4198 objects, 37,782 instruction-following pairs, and 47 categories. Experiments show that Building-MLLM achieves 88.00%, 65.10%, and 68.14% on the three task types, respectively, demonstrating superior indoor component language understanding and providing initial generalizability in transfer inference on other real-world datasets.

Visual-spatial understanding, defined as the ability to infer object relationships and scene layouts from visual inputs, is fundamental to downstream tasks such as robotic navigation and embodied interaction. However, pre-trained vision-language models (VLMs) remain constrained by spatial uncertainty stemming from inherently 2D observations and by the scarcity of data for 3D spatial understanding. To address these limitations, we proposed a novel framework, SpaceEra, in the NeurIPS 2025 Spotlight paper. Although it achieved significant performance gains, we further observed that its effectiveness is hindered by insufficient input from scanning videos and weak reasoning constraints. To tackle these newly emerged challenges, we extend the original framework into a comprehensive system, termed SpaceEra++, which spans data construction, model design, training optimization, and prompting inference. Specifically, to alleviate input insufficiency, we introduce ScenePick, a frame sampling strategy that balances spatial coverage with object semantics to produce compact yet comprehensive scene representations. In addition, to enhance spatial reasoning, we develop SpaceAlign, which enforces pairwise object constraints by jointly exploiting absolute coordinates and relative spatial relations, thereby aligning optimization with spatial accuracy. Extensive experiments across multiple benchmarks demonstrate consistent improvements over strong baselines, while ablation studies validate both the individual and joint contributions of each component, and further analyses provide guidance for future research.

In this paper, we study depth perception of vision-language models (VLMs) to isolate the effects of pictorial depth cues and disentangle vision and language influences on model performance. To this end, we combine depth-ordering and odd-one-out psychophysical tasks: the VLMs are presented with images where one object is at different depth relative to other, otherwise identical, objects, and must determine whether the odd-one-out target is closer or farther to the observer. To create stimuli, we generate 2D views from simulated and real 3D scenes while controlling the presence of individual pictorial depth cues, enabling a fine-grained analysis of cue-level contributions. Language effects are examined by varying referring expression clarity. We also introduce a novel metric to quantify vision-vs-language sensitivities. Applying this methodology, we create the Odd-One-Out Depth (O3-D) dataset with 37K real and synthetic images and 147K image-question pairs. Evaluation of 12 open-source and commercial models on O3-D shows under-utilization of depth cues and depth-ordering accuracies between 47% and 56%, with no model above chance level. At the same time, our metric reveals strong linguistic bias in the answers. Neither chain-of-thought (CoT) nor in-context learning (ICL) significantly improves performance, suggesting that static image data alone may be insufficient for depth understanding. All code, the image generation pipeline, and the O3-D dataset are publicly released at https://github.com/lyiqian/o3-d.

Identifying and grounding precise geometric entities, such as edges, planar regions, and curved surfaces within 3D objects, is foundational to computer-aided design (CAD), robotic manipulation, and scientific simulation. Although modern Vision Language Models (VLMs) have advanced referring segmentation (RIS) in the image domain, extending such language-driven localization to structured 3D geometry is substantially harder. The 3D object appearance is highly sensitive to viewpoints; a single perspective may render a target entity clearly observable, while another may suffer from severe occlusion or foreshortening. In this work, we attempt to solve these challenges with MV-GEL (Multi-View Geometric Entity Localization), a framework for localizing fine-grained geometric entities on polygon meshes from natural language queries. Our key insight is that reliable CAD entity (i.e., faces, edges or solids) localization depends on selecting views that make the queried entity maximally interpretable. We introduce GELviews, a prompt-conditioned ranking module that prioritizes viewpoints based on language prompted observability of geometric CAD entities. Selected views are processed by a VLM-based reasoning segmentation backbone, and predicted masks are lifted to the corresponding meshes via geometry-aware ray casting. Our framework is completely CAD agnostic and relies only on 3D meshes. Experiments show up to a 1.7X improvement in face-level IoU and over 4.5X gains in edge-level F1 compared to vanilla baselines, substantially outperforming CLIP-based and random view sampling, particularly for thin and view-sensitive structures.The dataset, code and trained checkpoints are available at https://github.com/kbali1297/MV-GEL.

3D Visual Grounding (3DVG) aims to localize target objects in 3D scenes given natural language descriptions. Existing approaches typically perform reasoning over the entire scene, leading to ambiguous predictions and high computational cost, especially in cluttered environments. We observe that many referential expressions rely on local spatial context and often correspond to restricted spatial regions rather than the full scene. Motivated by this insight, we propose PruneGround, an effective plug-and-play framework for 3DVG built upon three key components. First, we introduce Language-Guided Spatial Pruning (LGSP), which leverages a frozen Vision Language Model (VLM) to identify language-relevant regions, thereby reducing spatial computation and grounding candidates in the narrower search space. Second, we propose MultiView-Conditioned Description Reformulation (MCDR), which decomposes complex expressions into simplified target-anchor relations and augments missing spatial cues through multi-view reasoning. Finally, we propose LLM-Grounder, which repurposes a detection-pretrained spatial LLM into a language-conditioned grounding model by aligning point cloud and linguistic representations within the pruned region. Extensive experiments on the three most popular point cloud benchmarks demonstrate that our method achieves state-of-the-art results on all three ScanRefer settings and on 9 out of 10 Nr3D/Sr3D settings. Code and models are publicly available: https://github.com/leduckhai/PruneGround

Open-vocabulary 3D semantic Gaussian field learning relies on multi-view 2D supervision, whose semantic targets and spatial assignments are often unreliable. Across varying viewpoints, view-dependent features cause semantic identity drift, while propagated tracker masks introduce boundary leakage and identity switches. Directly optimizing against these unreliable 2D targets forces the 3D representation to absorb multi-view contradictions, leading to severe error accumulation. To resolve this limitation, we propose SAD-GS, a framework for learning reliable 3D semantic Gaussian fields via dynamic geo-semantic anchoring. Specifically, Semantic Anchor Distillation (SAD) distills per-view visual embeddings into consensus text anchors to establish a viewpoint-invariant semantic identity. Concurrently, the Geo-Semantic Feedback Loop (GSFL) leverages the evolving 3D field to actively filter tracker anomalies and refine spatial mask assignments via a conservative three-gate update rule. Extensive evaluations on LERF-OVS, 3D-OVS, and Mip-NeRF360 show that SAD-GS consistently achieves the best overall performance in both open-vocabulary localization and semantic segmentation. These comprehensive improvements validate the effectiveness and robustness of dynamic geo-semantic anchoring for reliable 3D semantic Gaussian field learning.

Vision-language pre-training (VLP) holds great promise for general-purpose medical AI by leveraging radiology reports as rich textual supervision, yet existing methods struggle with 3D CT imaging due to inefficient visual backbones and coarse semantic alignment. To address these issues, we propose a tailored VLP framework featuring three key components: (1) a CNN-ViT hybrid encoder that replaces ViT's patch embedding with a 3D CNN backbone to efficiently capture local anatomical details while preserving global attention and compatibility with pre-trained cross-modal priors; (2) a disease-level contrastive learning mechanism using learnable query tokens to dynamically extract disease-specific semantics from full reports and align them with corresponding visual features, thereby disentangling distinct diseases within the same anatomical region; and (3) a diagnosis-aware prompt strategy that employs real clinical phrases and aggregated disease prototypes to bridge the pre-training-inference gap and enhance zero-shot diagnostic reliability. Our model achieves state-of-the-art performance on CT-RATE (84.4% AUC, +5.1%) and Rad-ChestCT (75.4% AUC, +5.4%), with even larger gains (+9.8% AUC) on a challenging 60-disease benchmark, and demonstrates strong transferability to radiology report generation, underscoring the generality and clinical utility of our approach.

Vision-language contrastive pretraining has become the dominant recipe for 3D medical foundation models, leveraging the large volumes of paired scans and reports produced in clinical practice. However, medical images usually span dozens of organs, and radiological reports are much longer than typical natural image captions and are composed of multiple structured sections. CLIP-style pretraining compresses this structure by encoding each modality into a single global token, at the risk of losing important details. We introduce ConQuer (Concept Queries), an image-text pretraining method that augments CLIP's global alignment with a set of localized alignments, one per concept. ConQuer splits the report into concept-specific sections and learns cross-attention queries that pool the matching image features without using any segmentation mask or spatial supervision. Contrastive learning is then applied independently for each concept. Concepts can be any unit of semantic localization; here, they are anatomical regions, one query per organ or gross body region. As a byproduct, each query learns attention maps focused on its concept, providing built-in spatial interpretability. We use ConQuer to train Jolia, a 3D CT foundation model on chest and abdominal CT. Jolia consistently outperforms a CLIP baseline on findings classification, report generation, and cross-center transfer, and sets a new state of the art across multiple public benchmarks. Jolia's weights are available at https://huggingface.co/raidium/Jolia