A QA generation pipeline with True-Motion Tracking creates the 4DP-QA dataset and benchmark to improve VLM performance on motion-related 4D perception tasks.
hub Canonical reference
VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction
Canonical reference. 71% of citing Pith papers cite this work as background.
46
Pith papers citing it
Background
71% of classified citations
abstract
The rapid advancement of Large Multimodal Models (LMMs) for 2D images and videos has motivated extending these models to understand 3D scenes, aiming for human-like visual-spatial intelligence. Nevertheless, achieving deep spatial understanding comparable to human capabilities poses significant challenges in model encoding and data acquisition. Existing methods frequently depend on external depth sensors for geometry capture or utilize off-the-shelf algorithms for pre-constructing 3D maps, thereby limiting their scalability, especially with prevalent monocular video inputs and for time-sensitive applications. In this work, we introduce VLM-3R, a unified framework for Vision-Language Models (VLMs) that incorporates 3D Reconstructive instruction tuning. VLM-3R processes monocular video frames by employing a geometry encoder to derive implicit 3D tokens that represent spatial understanding. Leveraging our Spatial-Visual-View Fusion and over 200K curated 3D reconstructive instruction tuning question-answer (QA) pairs, VLM-3R effectively aligns real-world spatial context with language instructions. This enables monocular 3D spatial assistance and embodied reasoning. To facilitate the evaluation of temporal reasoning, we introduce the Vision-Spatial-Temporal Intelligence benchmark, featuring over 138.6K QA pairs across five distinct tasks focused on evolving spatial relationships. Extensive experiments demonstrate that our model, VLM-3R, not only facilitates robust visual-spatial reasoning but also enables the understanding of temporal 3D context changes, excelling in both accuracy and scalability.
hub tools
citation-role summary
background 10
baseline 2
method 2
citation-polarity summary
representative citing papers
Astra couples an RL-trained VLM policy with a view-consistent Bagel-based world simulator to enable agentic imagination during spatial reasoning, yielding benchmark gains on MMSI-Bench and MindCube.
Proposes Spatial Narrative Score (SNS) evaluation for VLMs' camera motion understanding and introduces CaMo model achieving consistent performance on SNS and direct QA.
Proposes an equation-anchored tool-use method for MLLMs that writes the pinhole back-projection equation in Chain-of-Thought and substitutes retrieved camera intrinsics and depths to achieve robustness in 3D object detection and visual grounding under rescaled intrinsics.
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
ViSRA boosts MLLM 3D spatial reasoning performance by up to 28.9% on unseen tasks via a plug-and-play video-based agent that extracts explicit spatial cues from expert models without any post-training.
4DThinker enables VLMs to perform dynamic spatial reasoning by thinking with 4D latent mental imagery using new fine-tuning and reinforcement learning methods.
A 3D-grounded autoencoder and diffusion transformer allow direct generation of 3D scenes in an implicit latent space using a fixed 1K-token representation for arbitrary views and resolutions.
EgoTL provides a new egocentric dataset with think-aloud chains and metric labels that benchmarks VLMs on long-horizon tasks and improves their planning, reasoning, and spatial grounding after finetuning.
Backward token warping in ViT-based MLLMs enables reliable reasoning from nearby viewpoints by preserving semantic coherence better than pixel-wise warping or fine-tuning baselines.
GeoThinker enables active, task-conditioned geometry integration in MLLMs via spatial-grounded fusion and importance gating, reaching 72.6 on VSI-Bench.
SpatialMosaic introduces a 2M-pair multi-view QA dataset and 1M-pair benchmark for MLLMs on spatial reasoning under partial visibility, plus a hybrid baseline that integrates 3D reconstruction models as geometry encoders.
4D-RGPT uses perceptual 4D distillation to boost region-level 4D perception in multimodal LLMs and reports gains on existing and new video QA benchmarks.
POMA-3D learns self-supervised 3D scene representations from point maps and improves performance on geometric 3D tasks including navigation and scene retrieval.
DR-MV3D decomposes MV3D-VQA into global map construction, question-conditioned view planning, and egocentric grounding, supervised by global consistency and local trajectory rewards optimized via GRPO.
A self-evolving training loop that generates its own spatial QA data with executable code and difficulty feedback lifts Qwen3-VL-4B/8B to 62.7/63.3 on VSI-Bench using an order of magnitude less data.
MotionVLA converts short past video windows into compact trajectory-field tokens to supply motion-consistent evidence for vision-language-action robot policies, improving long-horizon manipulation.
Stream3D-VLM adds autoregressive streaming control, VSFI geometry integration, GAVC compression, and a 1M-pair benchmark to enable real-time 3D VLM performance that beats prior models on 29 online and offline tasks.
Reasmory turns 3D reconstruction into validated program-executable memory for VLMs, yielding 6-18% gains on spatial reasoning benchmarks over direct baselines.
Standard VLMs achieve expert-level 3D performance on depth estimation, pose estimation, and object understanding via three simple techniques without architecture changes or regression losses.
GASP injects geometric priors into VLMs via a deep-supervised correspondence head trained on video point correspondences and depth consistency, raising internal matching accuracy and delivering gains on spatial benchmarks without any 3D VQA data.
ProSR adds a Counterfactual Invariance Penalty and a Tail Drift Penalty to shape VLM reasoning trajectories for better visual dependence and stability on spatial tasks.
FOUND-IT constructs evolving task-driven 3D scene graphs with on-demand granularity from monocular cameras by augmenting foundation models, reporting 79% higher accuracy on a grounding benchmark and real-time Jetson deployment.
GeoWeaver performs token-adaptive geometric grounding on visual tokens from a multi-level bank prior to language modeling to support better spatio-temporal reasoning.
citing papers explorer
-
4DP-QA: Scalable QA for 4D Perception in Vision Language Models
A QA generation pipeline with True-Motion Tracking creates the 4DP-QA dataset and benchmark to improve VLM performance on motion-related 4D perception tasks.
-
Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators
Astra couples an RL-trained VLM policy with a view-consistent Bagel-based world simulator to enable agentic imagination during spatial reasoning, yielding benchmark gains on MMSI-Bench and MindCube.
-
CaMo: Camera Motion Grounded Evaluation and Training for Vision-Language Models
Proposes Spatial Narrative Score (SNS) evaluation for VLMs' camera motion understanding and introduces CaMo model achieving consistent performance on SNS and direct QA.
-
Towards Camera-Robust 3D Localization: Equation-Anchored Tool-Use for MLLMs
Proposes an equation-anchored tool-use method for MLLMs that writes the pinhole back-projection equation in Chain-of-Thought and substitutes retrieved camera intrinsics and depths to achieve robustness in 3D object detection and visual grounding under rescaled intrinsics.
-
ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
-
ViSRA: A Video-based Spatial Reasoning Agent for Multi-modal Large Language Models
ViSRA boosts MLLM 3D spatial reasoning performance by up to 28.9% on unseen tasks via a plug-and-play video-based agent that extracts explicit spatial cues from expert models without any post-training.
-
4DThinker: Thinking with 4D Imagery for Dynamic Spatial Understanding
4DThinker enables VLMs to perform dynamic spatial reasoning by thinking with 4D latent mental imagery using new fine-tuning and reinforcement learning methods.
-
Any 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale
A 3D-grounded autoencoder and diffusion transformer allow direct generation of 3D scenes in an implicit latent space using a fixed 1K-token representation for arbitrary views and resolutions.
-
EgoTL: Egocentric Think-Aloud Chains for Long-Horizon Tasks
EgoTL provides a new egocentric dataset with think-aloud chains and metric labels that benchmarks VLMs on long-horizon tasks and improves their planning, reasoning, and spatial grounding after finetuning.
-
Token Warping Helps MLLMs Look from Nearby Viewpoints
Backward token warping in ViT-based MLLMs enables reliable reasoning from nearby viewpoints by preserving semantic coherence better than pixel-wise warping or fine-tuning baselines.
-
Thinking with Geometry: Active Geometry Integration for Spatial Reasoning
GeoThinker enables active, task-conditioned geometry integration in MLLMs via spatial-grounded fusion and importance gating, reaching 72.6 on VSI-Bench.
-
SpatialMosaic: A Multiview VLM Dataset for Partial Visibility
SpatialMosaic introduces a 2M-pair multi-view QA dataset and 1M-pair benchmark for MLLMs on spatial reasoning under partial visibility, plus a hybrid baseline that integrates 3D reconstruction models as geometry encoders.
-
4D-RGPT: Toward Region-level 4D Understanding via Perceptual Distillation
4D-RGPT uses perceptual 4D distillation to boost region-level 4D perception in multimodal LLMs and reports gains on existing and new video QA benchmarks.
-
POMA-3D: The Point Map Way to 3D Scene Understanding
POMA-3D learns self-supervised 3D scene representations from point maps and improves performance on geometric 3D tasks including navigation and scene retrieval.
-
Dense Reward for Multi-View 3D Reasoning with Global Maps and Local Views
DR-MV3D decomposes MV3D-VQA into global map construction, question-conditioned view planning, and egocentric grounding, supervised by global consistency and local trajectory rewards optimized via GRPO.
-
Ouroboros-Spatial: Closing the Data-Model Loop for Spatial Reasoning
A self-evolving training loop that generates its own spatial QA data with executable code and difficulty feedback lifts Qwen3-VL-4B/8B to 62.7/63.3 on VSI-Bench using an order of magnitude less data.
-
MotionVLA: Injecting Geometric Motion into Vision-Language-Action Model
MotionVLA converts short past video windows into compact trajectory-field tokens to supply motion-consistent evidence for vision-language-action robot policies, improving long-horizon manipulation.
-
Stream3D-VLM: Online 3D Spatial Understanding with Incremental Geometry Priors
Stream3D-VLM adds autoregressive streaming control, VSFI geometry integration, GAVC compression, and a 1M-pair benchmark to enable real-time 3D VLM performance that beats prior models on 29 online and offline tasks.
-
Reasmory: 3D Reconstruction as Explicit Memory for VLMs Spatial Reasoning
Reasmory turns 3D reconstruction into validated program-executable memory for VLMs, yielding 6-18% gains on spatial reasoning benchmarks over direct baselines.
-
VLM3: Vision Language Models Are Native 3D Learners
Standard VLMs achieve expert-level 3D performance on depth estimation, pose estimation, and object understanding via three simple techniques without architecture changes or regression losses.
-
Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning
GASP injects geometric priors into VLMs via a deep-supervised correspondence head trained on video point correspondences and depth consistency, raising internal matching accuracy and delivering gains on spatial benchmarks without any 3D VQA data.
-
ProSR: Process-Shaped Spatial Reasoning for Reliable Chain-of-Thought in VLMs
ProSR adds a Counterfactual Invariance Penalty and a Tail Drift Penalty to shape VLM reasoning trajectories for better visual dependence and stability on spatial tasks.
-
FOUND-IT: Foundation-model-first Task-driven 3D Scene Graphs with Granularity on Demand
FOUND-IT constructs evolving task-driven 3D scene graphs with on-demand granularity from monocular cameras by augmenting foundation models, reporting 79% higher accuracy on a grounding benchmark and real-time Jetson deployment.
-
GeoWeaver: Grounding Visual Tokens with Geometric Evidence before Scene Reasoning
GeoWeaver performs token-adaptive geometric grounding on visual tokens from a multi-level bank prior to language modeling to support better spatio-temporal reasoning.
-
Unlocking Dense Metric Depth Estimation in VLMs
DepthVLM converts a standard VLM into a dense metric depth predictor by attaching a lightweight head and training under unified vision-text supervision, outperforming prior VLMs and some pure vision models on a new indoor-outdoor benchmark.
-
$M^2$-VLA: Boosting Vision-Language Models for Generalizable Manipulation via Layer Mixture and Meta-Skills
Freezing a VLM backbone and routing its layers through Mixture-of-Layers plus a Meta-Skill memory yields higher success and stronger zero-shot generalization than fine-tuned VLAs on LIBERO and real robots.
-
Text-Guided 6D Object Pose Rearrangement via Closed-Loop VLM Agents
Closed-loop VLM agents with multi-view feedback, object axes overlays, and single-axis rotations predict text-guided goal 6D poses better than prior methods and improve zero-shot robot manipulation.
-
Lifting Unlabeled Internet-level Data for 3D Scene Understanding
Unlabeled web videos processed by designed data engines generate effective training data that yields strong zero-shot and finetuned performance on 3D detection, segmentation, VQA, and navigation.
-
MiMo-Embodied: X-Embodied Foundation Model Technical Report
MiMo-Embodied is a single foundation model that achieves state-of-the-art results on 17 embodied AI benchmarks and 12 autonomous driving benchmarks through multi-stage learning, curated data, and CoT/RL fine-tuning that produces positive cross-domain transfer.
-
Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modeling
Geometry Forcing aligns video diffusion representations with geometric foundation model features via angular cosine and scale regression objectives to improve 3D consistency in generated videos.
-
HoloAgent-0: A Unified Embodied Agent Framework with 3D Spatial Memory
HoloAgent-0 is a unified embodied agent framework with Embodied AgentOS, 3D spatial memory, and embodied skills, deployed and evaluated on real robot hardware for navigation and manipulation tasks.
-
Learning Geometric Representations from Videos for Spatial Intelligent Multimodal Large Language Models
GeoVR distills camera pose, depth, scale, and multi-scale 3D features from pre-trained models into MLLMs via video supervision to improve spatial reasoning.
-
LongSpace: Exploring Long-Horizon Spatial Memory from Perception to Recall in Video
Presents LongSpace-Bench benchmark and LongSpace framework that chunks long videos, adds 3D structural cues, and builds layer-aware memory to improve spatial reasoning in multimodal LLMs.
-
GEM: Generative Supervision Helps Embodied Intelligence
GEM adds generative depth supervision to VLM pre-training and reports improved results on embodied benchmarks plus real-world robot execution.
-
Q-GeoMem: Question-Guided Geometric Memory for Video Spatial Reasoning
Question-guided dual geometric memories with relevance-novelty utility reportedly reach state-of-the-art video spatial reasoning on two in-domain and five out-of-distribution benchmarks.
-
Rethinking VLM Representation for VLA Initialization
Experiments indicate original VLM representations are crucial for VLA performance, LoRA outperforms full finetuning, and staged robot-data pretraining yields the strongest initialization.
-
Thinking with Novel Views: A Systematic Analysis of Generative-Augmented Spatial Intelligence
Integrating generative novel-view synthesis into LMM reasoning loops improves accuracy on spatial subtasks by 1.3 to 3.9 percentage points across multiple models and tasks.
-
From Where Things Are to What They Are For: Benchmarking Spatial-Functional Intelligence in Multimodal LLMs
SFI-Bench shows current multimodal LLMs struggle to integrate spatial memory with functional reasoning and external knowledge in video tasks.
-
SpatialImaginer: Towards Adaptive Visual Imagination for Spatial Reasoning
SpatialImaginer integrates visual imagination with textual chain-of-thought to improve spatial reasoning robustness in multimodal large language models.
-
FF3R: Feedforward Feature 3D Reconstruction from Unconstrained views
FF3R unifies geometric and semantic 3D reconstruction in a single annotation-free feed-forward network trained solely via RGB and feature rendering supervision.
-
MAG-3D: Multi-Agent Grounded Reasoning for 3D Understanding
MAG-3D is a training-free multi-agent framework that coordinates planning, grounding, and coding agents with off-the-shelf VLMs to achieve grounded 3D reasoning and state-of-the-art benchmark results.
-
OpenSpatial: A Principled Data Engine for Empowering Spatial Intelligence
OpenSpatial supplies a principled open-source data engine and 3-million-sample dataset that raises spatial-reasoning model performance by an average of 19 percent on benchmarks.
-
EgoMind: Activating Spatial Cognition through Linguistic Reasoning in MLLMs
EgoMind uses Role-Play Caption and Progressive Spatial Analysis to give MLLMs competitive multi-frame spatial reasoning without 3D priors, using only 5K SFT and 20K RL samples.
-
AVA-VLA: Improving Vision-Language-Action models with Active Visual Attention
AVA-VLA reformulates VLA learning as a POMDP using recurrent states and active visual attention to achieve state-of-the-art results on LIBERO, CALVIN, and real dual-arm tasks.
-
Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models
An 8B embodied foundation model trained on 15B tokens with multi-task RL and a Planner-Grounder-Corrector loop claims SOTA on 16/24 embodied VLM benchmarks and strong VLA/real-robot transfer.
- Let Geometry GUIDE: Layer-wise Unrolling of Geometric Priors in Multimodal LLMs