@_akhaliq: Representing the Unseen Rethinking Visual Imagination for Spatial Reasoning paper: https://huggingface.co/papers/2610.3…
Summary
The paper introduces ImagiRep, a simulator-grounded benchmark showing that spatial reasoning benefits from diverse representations beyond RGB, and proposes GEARS, a lightweight test-time strategy for adaptively selecting representations to improve reasoning about unobserved states.
View Cached Full Text
Cached at: 10/28/26, 07:43 PM
Representing the Unseen
Rethinking Visual Imagination for Spatial Reasoning
paper: https://t.co/SgjDTxldcm https://t.co/jILqFiwrp1
Paper page - Representing the Unseen: Rethinking Visual Imagination for Spatial Reasoning
Source: https://huggingface.co/papers/2610.37437
Abstract
ImagiRep reveals that spatial reasoning benefits from diverse representations beyond RGB, and GEARS adaptively selects representations to improve reasoning about unobserved states.
Visual world modelsincreasingly supportspatial reasoningby generatingRGBviews of unobserved states. However,RGBhas become the default interface largely because it is well suited for visual generation, rather than necessarily being the most effective representation for downstream reasoning. Different spatial questions may benefit from different representations of the same imagined state:top-down geometrycan facilitate global distance comparisons, whilepanoramic viewsprovide broader context for reasoning about open space. To isolate the role of representation in spatial imagination, we introduceImagiRep, a simulator-grounded benchmark of 922 questions spanning seven question types across two settings, Viewpoint Imagination andCross-view Integration, providing oracleRGBand six alternative representations (e.g.,depth,panoramic views, andtop-down geometry) of the same unobserved state. This controlled comparison reveals that even accurate imaginedRGBdoes not guarantee accuratespatial reasoning, and different questions benefit from different representations. Similar trends hold for representations derived from world-model-estimated states. Notably, estimatedRGBalone can misleadspatial reasoning, while alternative representations retain complementary benefits. Building on these findings, we introduceGEARS, a lightweight test-time strategy foradaptive representation selection, outperforming existing spatial imagination methods and showing that reasoning about unseen states requires more than generatingRGBviews.
View arXiv pageView PDFAdd to collection
Get this paper in your agent:
hf papers read 2610\.37437
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2610.37437 in a model README.md to link it from this page.
Datasets citing this paper1
#### Imagirep/ImagiRep Viewer• Updatedabout 5 hours ago • 922
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2610.37437 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators
The paper proposes Astra, an agentic spatial reasoning framework that couples a reinforcement learning-trained VLM policy with a world simulator to generate novel-view observations for improved spatial reasoning in Vision-Language Models.
Retrieve, Integrate, and Synthesize: Spatial-Semantic Grounded Latent Visual Reasoning
This paper introduces RIS, a framework for spatial-semantic grounded latent visual reasoning in Multimodal Large Language Models to overcome information bottlenecks. It proposes anchoring latent tokens to spatial and semantic evidence, showing improvements on benchmarks like V* and HRBench.
Reason, Then Re-reason: Cross-view Revisiting Improves Spatial Reasoning
A training-free framework for spatial reasoning from egocentric videos that enables revisiting conclusions through synthesized novel-view videos generated from predicted 3D geometry.
Thinking with Visual Grounding
This paper introduces visually grounded thinking, a method for vision-language models to interleave natural-language reasoning with explicit visual evidence grounding using points or boxes. A scalable synthesis pipeline and grounding-aware reinforcement learning improve reasoning accuracy, enabling a 4B model to match or surpass a 27B model on spatial and counting benchmarks.
SpatialASM: Internalizing Active Mentalization for Spatial Reasoning
The paper introduces SpatialASM, a framework that trains large vision-language models to actively plan task-relevant viewpoints and mentally imagine missing spatial evidence via novel-view synthesis scaffolding, improving spatial reasoning across four backbones and seven benchmarks without extra views at inference.