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.
hub Baseline reference
Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces
Baseline reference. 70% of citing Pith papers use this work as a benchmark or comparison.
38
Pith papers citing it
Baseline
70% of classified citations
abstract
Humans possess the visual-spatial intelligence to remember spaces from sequential visual observations. However, can Multimodal Large Language Models (MLLMs) trained on million-scale video datasets also ``think in space'' from videos? We present a novel video-based visual-spatial intelligence benchmark (VSI-Bench) of over 5,000 question-answer pairs, and find that MLLMs exhibit competitive - though subhuman - visual-spatial intelligence. We probe models to express how they think in space both linguistically and visually and find that while spatial reasoning capabilities remain the primary bottleneck for MLLMs to reach higher benchmark performance, local world models and spatial awareness do emerge within these models. Notably, prevailing linguistic reasoning techniques (e.g., chain-of-thought, self-consistency, tree-of-thoughts) fail to improve performance, whereas explicitly generating cognitive maps during question-answering enhances MLLMs' spatial distance ability.
hub tools
citation-role summary
dataset 5
background 3
baseline 2
citation-polarity summary
representative citing papers
PInVerify is a new offline embodied benchmark for active instance verification that supplies multi-view captures and 6-sector navigation topology, with MLLM baselines reaching 85.6% after fine-tuning but showing no reliable benefit from tested next-best-view strategies.
SpaceDG is the first large-scale benchmark dataset (~1M QA pairs) simulating nine visual degradations in 3DGS-rendered scenes to measure and improve spatial intelligence robustness in MLLMs.
The Token Replacement Test shows VLMs keep most accuracy gains even after corrupting or replacing continuous thought token content, indicating the tokens are not used as information bottlenecks.
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.
Video-Holmes benchmark shows top MLLMs achieve at most 45% accuracy on tasks needing integration of multiple clues from suspense films, unlike existing perception-focused tests.
SpaceR uses a new verifiable dataset and map-imagination-augmented RLVR to reach SOTA spatial reasoning accuracy in MLLMs, exceeding GPT-4o on VSI-Bench.
Video-R1 uses temporal-aware RL and mixed datasets to boost video reasoning in MLLMs, with a 7B model reaching 37.1% on VSI-Bench and surpassing GPT-4o.
EFlow improves long-video QA by separating clip-finding from answering and adding a confidence trigger that re-reads the full video when unsure.
SatAgent is a UAV-satellite collaborative spatial reasoning model using geometric 3D encoding, multi-view alignment, and a new 130K dataset that reports 25.91% and 11.69% gains over general and specialized baselines.
A 9B multimodal model learns to tailor raw video/GUI streams into schema-aligned training data, matching a proprietary annotator on downstream tasks; the abstract's capacity-scaling claims are not supported by the body.
S-Agent augments VLMs with spatial tools, scene and agent memory for evidence accumulation on multi-view and video tasks, and produces an 8B model via SFT on its own trajectories that beats same-scale baselines.
IPT supervision improves spatial reasoning in VLMs on perspective taking, path tracing, and multiview counting tasks, often outperforming textual chain-of-thought while remaining consistent with observed inputs.
MOSS-Video-Preview introduces a cross-attention architecture and synthesized real-time QA data to enable continuous perception, answer revision, and faster inference in video-language models compared to decoder-only designs.
SOCO is a new benchmark for semantic object correspondence that provides taxonomy, annotations, and language labels to evaluate part-level understanding in vision and multimodal foundation models.
Multi-SpatialMLLM integrates depth perception, visual correspondence, and dynamic perception into MLLMs via a 27M-sample MultiSPA dataset and benchmark, yielding gains on multi-frame spatial tasks.
Flat-Pack Bench is a new evaluation suite that shows state-of-the-art LVLMs perform poorly on nuanced spatio-temporal reasoning required for furniture assembly videos.
SaaS-Bench benchmark shows LLM-based agents achieve under 4% end-to-end success on 106 realistic professional tasks spanning 23 deployable SaaS platforms.
SpaceMind++ adds an explicit voxelized allocentric cognitive map and coordinate-guided fusion to video MLLMs, claiming SOTA on VSI-Bench and improved out-of-distribution generalization on three other 3D benchmarks.
Video-ToC adds tree-guided cue localization, demand-based RL rewards, and automated datasets to video LLMs, reporting better results than prior methods on six understanding benchmarks plus a hallucination test.
InternVL3.5 advances open-source multimodal models with Cascade RL for +16% reasoning gains and ViR for 4x inference speedup, with the 241B model reaching SOTA among open-source MLLMs on multimodal, reasoning, and agentic tasks.
Embodied-R1 uses a pointing-centric representation and reinforced fine-tuning on a 200K dataset to achieve state-of-the-art results on embodied benchmarks plus 56.2% success in SIMPLEREnv and 87.5% on real XArm tasks without task-specific training.
VILASR integrates visual drawing operations with reasoning in LVLMs via cold-start synthetic training, reflective rejection sampling, and reinforcement learning, yielding an 18.4% average gain on spatial reasoning benchmarks.
Spatial-MLLM adds a 3D spatial encoder initialized from a visual geometry model and space-aware frame sampling to MLLMs to improve spatial understanding and reasoning from purely 2D visual inputs.
citing papers explorer
-
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.
-
PInVerify: An Offline Embodied Benchmark for Active Instance Verification
PInVerify is a new offline embodied benchmark for active instance verification that supplies multi-view captures and 6-sector navigation topology, with MLLM baselines reaching 85.6% after fine-tuning but showing no reliable benefit from tested next-best-view strategies.
-
SpaceDG: Benchmarking Spatial Intelligence under Visual Degradation
SpaceDG is the first large-scale benchmark dataset (~1M QA pairs) simulating nine visual degradations in 3DGS-rendered scenes to measure and improve spatial intelligence robustness in MLLMs.
-
Ablate-to-Validate: Are Vision-Language Models Really Using Continuous Thought Tokens?
The Token Replacement Test shows VLMs keep most accuracy gains even after corrupting or replacing continuous thought token content, indicating the tokens are not used as information bottlenecks.
-
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.
-
Video-Holmes: Can MLLM Think Like Holmes for Complex Video Reasoning?
Video-Holmes benchmark shows top MLLMs achieve at most 45% accuracy on tasks needing integration of multiple clues from suspense films, unlike existing perception-focused tests.
-
SpaceR: Reinforcing MLLMs in Video Spatial Reasoning
SpaceR uses a new verifiable dataset and map-imagination-augmented RLVR to reach SOTA spatial reasoning accuracy in MLLMs, exceeding GPT-4o on VSI-Bench.
-
Video-R1: Reinforcing Video Reasoning in MLLMs
Video-R1 uses temporal-aware RL and mixed datasets to boost video reasoning in MLLMs, with a 7B model reaching 37.1% on VSI-Bench and surpassing GPT-4o.
-
EFlow: Learning Evidence Flow for Long-Video Reasoning with Adaptive Reflection
EFlow improves long-video QA by separating clip-finding from answering and adding a confidence trigger that re-reads the full video when unsure.
-
AeroVerse-SatAgent: UAV-Satellite Collaborative Spatial Reasoning Inspired by the Dual Visual Pathway Theory of Cognitive Neuroscience
SatAgent is a UAV-satellite collaborative spatial reasoning model using geometric 3D encoding, multi-view alignment, and a new 130K dataset that reports 25.91% and 11.69% gains over general and specialized baselines.
-
DataClaw0: Agentic Tailoring Multimodal Data from Raw Streams
A 9B multimodal model learns to tailor raw video/GUI streams into schema-aligned training data, matching a proprietary annotator on downstream tasks; the abstract's capacity-scaling claims are not supported by the body.
-
S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence
S-Agent augments VLMs with spatial tools, scene and agent memory for evidence accumulation on multi-view and video tasks, and produces an 8B model via SFT on its own trajectories that beats same-scale baselines.
-
Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models
IPT supervision improves spatial reasoning in VLMs on perspective taking, path tracing, and multiview counting tasks, often outperforming textual chain-of-thought while remaining consistent with observed inputs.
-
MOSS-Video-Preview: Toward Real-Time Video Understanding via Cross-Attention
MOSS-Video-Preview introduces a cross-attention architecture and synthesized real-time QA data to enable continuous perception, answer revision, and faster inference in video-language models compared to decoder-only designs.
-
SOCO: Benchmarking Semantic Object Correspondence in Vision Foundation Models
SOCO is a new benchmark for semantic object correspondence that provides taxonomy, annotations, and language labels to evaluate part-level understanding in vision and multimodal foundation models.
-
Multi-SpatialMLLM: Multi-Frame Spatial Understanding with Multi-Modal Large Language Models
Multi-SpatialMLLM integrates depth perception, visual correspondence, and dynamic perception into MLLMs via a 27M-sample MultiSPA dataset and benchmark, yielding gains on multi-frame spatial tasks.
-
Flat-Pack Bench: Evaluating Spatio-Temporal Understanding in Large Vision-Language Models through Furniture Assembly
Flat-Pack Bench is a new evaluation suite that shows state-of-the-art LVLMs perform poorly on nuanced spatio-temporal reasoning required for furniture assembly videos.
-
SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?
SaaS-Bench benchmark shows LLM-based agents achieve under 4% end-to-end success on 106 realistic professional tasks spanning 23 deployable SaaS platforms.
-
SpaceMind++: Toward Allocentric Cognitive Maps for Spatially Grounded Video MLLMs
SpaceMind++ adds an explicit voxelized allocentric cognitive map and coordinate-guided fusion to video MLLMs, claiming SOTA on VSI-Bench and improved out-of-distribution generalization on three other 3D benchmarks.
-
Video-ToC: Video Tree-of-Cue Reasoning
Video-ToC adds tree-guided cue localization, demand-based RL rewards, and automated datasets to video LLMs, reporting better results than prior methods on six understanding benchmarks plus a hallucination test.
-
InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
InternVL3.5 advances open-source multimodal models with Cascade RL for +16% reasoning gains and ViR for 4x inference speedup, with the 241B model reaching SOTA among open-source MLLMs on multimodal, reasoning, and agentic tasks.
-
Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation
Embodied-R1 uses a pointing-centric representation and reinforced fine-tuning on a 200K dataset to achieve state-of-the-art results on embodied benchmarks plus 56.2% success in SIMPLEREnv and 87.5% on real XArm tasks without task-specific training.
-
Reinforcing Spatial Reasoning in Vision-Language Models with Interwoven Thinking and Visual Drawing
VILASR integrates visual drawing operations with reasoning in LVLMs via cold-start synthetic training, reflective rejection sampling, and reinforcement learning, yielding an 18.4% average gain on spatial reasoning benchmarks.
-
Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence
Spatial-MLLM adds a 3D spatial encoder initialized from a visual geometry model and space-aware frame sampling to MLLMs to improve spatial understanding and reasoning from purely 2D visual inputs.
-
Grounded Reinforcement Learning for Visual Reasoning
ViGoRL introduces visually grounded RL that anchors reasoning steps to image coordinates and uses multi-turn zooming to outperform standard RL and supervised baselines on spatial and GUI reasoning benchmarks.
-
VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction
VLM-3R augments VLMs with implicit 3D tokens from monocular video via geometry encoding and 200K+ 3D reconstructive QA pairs, plus a new 138K-pair temporal benchmark, to support spatial and embodied reasoning.
-
InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
InternVL3-78B sets a new open-source SOTA of 72.2 on MMMU via native joint multimodal pre-training, V2PE, MPO, and test-time scaling while remaining competitive with proprietary models.
-
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.
-
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.
-
SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model
SpatialVLA adds 3D-aware position encoding and adaptive discretized action grids to visual-language-action models, enabling strong zero-shot performance and fine-tuning on new robot setups after pre-training on 1.1 million real-world episodes.
-
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.
-
LASAR: Towards Spatio-temporal Reasoning with Latent Cognitive Map
LASAR pairs a dual-memory system with spatio-temporal contrastive learning to induce latent cognitive maps, reporting 2-3.5% zero-shot gains on VLN-CE and VSI-Bench plus high map self-consistency.
-
EasyVideoR1: Easier RL for Video Understanding
EasyVideoR1 delivers an optimized RL pipeline for video understanding in large vision-language models, achieving 1.47x throughput gains and aligned results on 22 benchmarks.
-
Vision-EKIPL: External Knowledge-Infused Policy Learning for Visual Reasoning
Vision-EKIPL injects high-quality actions from external models into RL training to expand exploration and raise the reasoning ceiling of MLLMs, reporting up to 5% gains on the Reason-RFT-CoT benchmark.
-
SpaceEra++: A Unified Framework Towards 3D Spatial Reasoning in Video
SpaceEra++ adds ScenePick frame sampling and SpaceAlign pairwise constraints to the prior SpaceEra system, claiming consistent benchmark gains for 3D video spatial reasoning.
-
Answer Self-Consistency with Margin-Triggered Question Re-Arbitration for the CVPR 2026 VidLLMs Challenge
ASC-MQRA applies answer self-consistency across stochastic video QA runs and optional margin-triggered re-arbitration to achieve 81.16% average accuracy on the CVPR 2026 VidLLMs Challenge Track 2 test set.
-
Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey
The paper provides the first comprehensive survey of multimodal chain-of-thought reasoning, including foundational concepts, a taxonomy of methodologies, application analyses, challenges, and future directions.