Image created with Gemini. Image prompt: Vintage book jacket illustration of an Edwardian gentleman in silhouette from behind wearing round spectacles, holding a straw boater hat upturned in his hands, with a miniature pop-up diorama of Edwardian Dublin rising from inside the hat complete with tram tracks, a winding river, terraced rooftops and drifting chimney smoke, bold serif title text ‘World Models’ across the top of the composition, flat inked screen-print colors in Shakespeare-and-Company blue, cream, lemon yellow and slate gray with a warm brick accent, visible paper grain and soft overcast daylight, landscape 16:9.
[2606.17030] Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
https://arxiv.org/abs/2606.17030
📣 Introducing the Qwen-Robot Suite — Qwen-RobotNav, Qwen-RobotManip, Qwen-RobotWorld, three foundation models, a full stack for embodied intelligence. 🧭 Qwen-RobotNav — the gateway to mobility. • Unifies 5 navigation tasks in one model: instruction following, point-goal,
https://x.com/Alibaba_Qwen/status/2066870197122899980
City scale 3d gaussian splats just hit different – well done apple maps
https://x.com/bilawalsidhu/status/2066175946831348094
MeshFlow: Efficient Artistic Mesh Generation via MeshVAE and Flow-based Diffusion Transformer”” TL;DR: learns a continuous mesh latent space and generates vertices and connectivity in parallel with flow matching, producing quality 3D meshes up to 18× faster than autoregressive.
https://x.com/Almorgand/status/2065557982985719841
MolmoMotion: Language-guided 3D motion forecasting
https://huggingface.co/blog/allenai/molmomotion
A multimodal data capture rig built for Physical AI training: Synchronizing RGB video, stereo IR depth, real-time 3D hand reconstruction, and 23×20 tactile pressure grids from sensor gloves during a cucumber-slicing task. The video from Grace Zhang visualizes aligned streams:
https://x.com/IlirAliu_/status/2066437515150586008
A robotics world model predicting 3D motion traces of semantic points instead of pixels or latents. Semantic points? Like objects, hands, and contacts! The approach uses video-only pretraining via TraceExtract on human videos to build an interpretable “physical language” that
https://x.com/IlirAliu_/status/2066582302952448302
By treating natural language as a universal action interface，Qwen-RobotWorld bridges the gap between general video generation models and domain-specific embodied models — this converts end-effector poses, steering commands, and navigation waypoints into a single interface,
https://x.com/Alibaba_Qwen/status/2066870210716647591
I made Physical AutoResearch sound simple (conceptually), but it took a village to pull off and lots of design thinking into the robot /loopcraft. The hardest part is everything we need to setup *before* pressing Enter. Here’s a behind-the-scene tour: 1. Safety harness Letting
https://x.com/DrJimFan/status/2067283904986517866
World Tracing Generative Pixel-Aligned Geometry Beyond the Visible
https://x.com/_akhaliq/status/2066926594698907780
μ_0 A Scalable 3D Interaction-Trace World Model
https://x.com/_akhaliq/status/2066927000564978054
Our $310 Million Fundraise to Accelerate World Simulation
https://odyssey.ml/our-series-b
Leave a Reply