Filter
Exclude
Time range
-
Near
👁️ Imaginative Perception Tokens (IPT): VLMs learn spatial reasoning from unseen viewpoints & occluded spaces. Externalize what a model *would* perceive — 3D awareness from 2D alone. #AI #ComputerVision #VLM #SpatialReasoning
14
Model 2.0 can reason spatial logic and text rendering far better than standard models. This hyper realistic shot isn't just a texture paste. Observe the perfect refraction logic: the entire poster wall, including the text "TIME TAKES EVERYTHING" and "I WAS SOMEONE", is visible and inverted within the water on the street, obeying the laws of physics. Mastering surface textures and complex text logic. Profile: imagine.art/c/thoai66 @ImagineArt_X @imagineart_creo #ImagineArt2Creatathon #TextRendering #MacroPhotography #SpatialReasoning #ImagineArt2
The future of AI spatial reasoning is here. ImagineArt 2.0 isn't just generating pixels; it's understanding the laws of physics. In this hyper realistic macro shot of a Jewel Chameleon, observe the perfect refraction logic: the chameleon is visible and inverted within the water drops, acting as a natural convex lens. From the microscopic skin textures to the complex light behavior within the morning mist, this is what "Reasoning based AI" truly looks like. My Community Profile: imagine.art/c/thoai66 @ImagineArt_X @imagineart_creo #ImagineArt2Creatathon #AIArt #MacroPhotography #SpatialReasoning #ImagineArt2
4
103
The future of AI spatial reasoning is here. ImagineArt 2.0 isn't just generating pixels; it's understanding the laws of physics. In this hyper realistic macro shot of a Jewel Chameleon, observe the perfect refraction logic: the chameleon is visible and inverted within the water drops, acting as a natural convex lens. From the microscopic skin textures to the complex light behavior within the morning mist, this is what "Reasoning based AI" truly looks like. My Community Profile: imagine.art/c/thoai66 @ImagineArt_X @imagineart_creo #ImagineArt2Creatathon #AIArt #MacroPhotography #SpatialReasoning #ImagineArt2
$10,000 Prize Pool. 30 Winners. 10 Days! 🚨 THE IMAGINEART 2.0 CREATATHON IS LIVE. ImagineArt 2.0 doesn't just generate, it REASONS. Composition. Lighting. Materials. Scene logic. It thinks before it creates. Now it's your turn. Push it. Break it. Show the world what this model can really do.
4
17
418
EO-2 is the hardware. Geospatial reasoning is the software. Together they turn raw pixels into context and that context into decisive action. #ArzIntelligence #SpatialReasoning #EO2 #DataSovereignty #GeospatialIntelligence
3
1,018
3 for 3 👇🏼 #dpmath friends! 😉 ➡️ March "Creativity Counts in Math" giveaway winners: We hope your Ss are enjoying exercising their #spatialreasoning muscles w/ @DragonFjord's 'A-Puzzle-a-Day' ➡️ Our April CCM focuses on "impossible shapes" inspired by M.C. Escher. 💕 ENJOY!
1
2
4
492
VLMs often look spatially smart, but they fall back on 2D appearance shortcuts instead of real geometry. GeoSR makes geometry truly matter for spatial reasoning in both static and dynamic videos. 🏆 51.9 on VSI-Bench 🏆 66.1 on DSR-Bench ( 7.2) #GeoSR #VLM #SpatialReasoning
1
1
7
791
@ace_robotics, part of SenseTime’s innovative business, introduces ACE-Brain-0—a groundbreaking open-source foundation model that empowers cross-embodiment #SpatialReasoning across #robotics, #AutonomousDriving, and unmanned aerial vehicle (#UAV) systems.   This innovation is injecting significant momentum into the development of the embodied intelligence industry.
🌟@ace_robotics introduces ACE-Brain-0 — the first spatial-intelligence-based open-source foundation model designed to unify embodied intelligence across different physical embodiments.
1
3
215
If you're at #WACV2026, come visit our CVP poster! 📄 arxiv.org/pdf/2512.08135 [Poster session] 🗓️Sun, Mar 8, 2026 • 4:00 PM – 5:45 PM MST 📍Tucson Ballroom & Prefunction Space 84 Our authors will be there and are happy to chat about spatial reasoning, multimodal models, and vision-inspired architectures. 👋 @wacv_official @mlpcucsd @LambdaAPI #spatialReasoning #MultimodalModel #VLM #3dvision
1
3
405
🚀 Excited to share our #WACV2026 paper for 3D spatial reasoning: CVP: Central-Peripheral Vision-Inspired Multimodal Model for Spatial Reasoning Inspired by human vision, we introduce CVP, which combines: 👁️Target-affinity tokens (central vision) to focus on relevant objects 🌍Allocentric grids (peripheral vision) to capture global scene context This simple idea significantly improves 3D spatial reasoning, achieving SOTA performance across multiple benchmarks. 📄Paper: arxiv.org/pdf/2512.08135 🌐Page: zeyuan-chen.com/cvp/ #spatialReasoning #MultimodalModel #VLM @LambdaAPI @UCSD @mlpcucsd @wacv_official
5
75
3,903
GeoAI predicts what's likely. Geospatial reasoning asks what's possible. When infrastructure fails, the gap between pattern learning and structural understanding becomes clear. Full article: geoawesome.com/geospatial-re… #Geoawesome #GeoAI #SpatialReasoning #ClimateAdaptation #GIS
5
19
1,189
How LLMs are transforming Message Sequence Charts into spatio-temporal reasoning tools. At Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD 2025), our researchers, in collaboration with @iiit_hyderabad , presented two LLM-based approaches that add spatial knowledge to Message Sequence Charts (MSCs), transforming them into advanced tools for spatio-temporal reasoning in narratives. Know more- bit.ly/3O5XMkl #Research #LLM #SpatialReasoning
1
4
208
1/7 Spatial reasoning is the *blindspot* of vision-language models. We just released: “The Spatial Blindspot of Vision-Language Models” Why VLMs miss left/right, layouts, counting — and what design choices help. #VLM #Multimodal #SpatialReasoning
1
5
21
2,769
A dice is rested in a certain position. If the dice is rolled on the showed path in the image, which number will appear on the top? #LogicalReasoning #IQTest #SpatialReasoning
2
1
3
737
Jan 16
I used the @Yupp_ai comparison tool to test how LLMs handle 3D geometry. While the math was easy, the logic behind the "intersection" of clock hands revealed a massive hallucination. The Challenge: Calculate the angle at 9:15 and determine if the hands intersect in 3D space (given they are stacked on an axle). The Fail (DeepSeek V3.2 Thinking):The model claims that even if the hands were perfectly aligned in 2D, they still wouldn't intersect in 3D because they are "like two parallel lines offset in height." The Reality Check: Clock hands aren't floating independently in a vacuum. They are radial segments mounted on a shared central axle. Geometrically, any two segments sharing a vertex (the pivot point) must intersect at that vertex. Mechanically, saying they don't intersect is like saying the spokes of a wheel don't meet at the hub. DeepSeek completely ignored the "pivot" and treated a physical object as abstract, disconnected lines. A perfect example of why LLMs still struggle with the physical world! #StrawberrySeeds #Yupp #DeepSeek #AI #SpatialReasoning #LLM
1
8
66
👁️ One camera sees, four cameras understand. With #GMSL, reComputer #Robotics J4012 runs 4 synchronized cameras at the edge—powering real-time detection with #YOLOv11 and #3D #sceneunderstanding & #spatialreasoning with #VGGT—all running directly at the edge. 🔗 Discover more about reComputer Robotics: seeedstudio.com/reComputer-R…
1
19
145
11,009
View-centric space matters. Most image-to-3D models operate in canonical space. That works for instances — but not for scenes. 🔑 Key difference: Scene generation is about #spatialreasoning. In canonical space: • Left/right in the image becomes meaningless • Different views collapse to the same 3D layout • Fine for instances, bad for scene layout 👁️ View-centric space fixes this: 2D image structure now directly corresponds to 3D spatial layout. 💡 We find both are necessary: Scene context attention view-centric space Remove either just not work. 🧵 [3/6]
1
7
1,035
3rd and 4th graders @WilkersonElem had such a fun and challenging time with #pentominoes in the library last week. @JCPS_LMS @JCASLKY #jcpslibraries #jcaslky #problemsolvers #learningfun #spatialreasoning
1
4
124
AI that sees, reasons, and collaborates in complex environments. With Embodied-RAG YOWO, Fujitsu Research CMU bring that to life in FieldWorkArena. Smarter agents, safer workplaces, faster response. 🛠️ 👉 See how it works: blog-en.fltech.dev/entry/202… #AI #MachineLearning #SpatialReasoning #SmartFactories #Robotics #EdgeAI

ALT Hierarchical structure of Embodied-RAG’s semantic forest, enabling multi-resolution query handling

1
7
324
🚀 Model and data for our CubifyAnything project are now released! 🔗 github.com/apple/ml-cubifyan… #SpatialReasoning #3DObjectDetection #transformers #detection #ai #genai
1
4
282