🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Generalist robot policies should be evaluated on 'generalization' metrics rather than merely reporting success rate in performance in arbitrary scenarios. ★-Gen introduces axes of generalization across inputs/outputs of policies & guides eval benchmarks for robot policies. 1/7

Feature Splatting can now turn Objaverse assets into GS. With optimized kernel, 30K iters can be done in <1min on a single 4090 GPU.

We will be hosting the workshop on "Agents in Interactions, from Humans to Robots" at CVPR2025 #CVPR2025 , welcome to join us by submitting a paper or stopping by our talks/posters! For more info please check out: agents-in-interactions.github.io

Check out our new work on exploring 3D Spatial Understanding with Multimodal LLMs!🚀 📀CA-VQA: A fine-tuning dataset and benchmark w/ various input signals and spatial tasks. 🤖MM-Spatial: A generalist MLLM excelling at spatial reasoning. 🔗arxiv.org/abs/2503.13111 🧵(1/n)

Diverse training data leads to a more robust humanoid manipulation policy, but collecting robot demonstrations is slow. Introducing our latest work, Humanoid Policy ~ Human Policy. We advocate human data as a scalable data source for co-training egocentric manipulation policy.⬇️

[1/n] We are releasing M3 (#ICLR2025): a Gaussian Splatting method that builds LMM memories for arbitrary scenes. 🔥 [Efficient] 16 degrees in each Gaussian primitive for one LMM. 🔥 [Alignment] The rendered features are directly in the source LMM embedding space.

🔥 VLMs aren’t built for spatial reasoning — yet. They hallucinate free space. Misjudge object fit. Can’t tell below from behind We built RoboSpatial to tackle that — a dataset for teaching spatial understanding to 2D/3D VLMs for robotics. 📝 Perfect review scores #CVPR2025 2025

Kids in 2025: writing Tom & Jerry instead of watching them. TTT + diffusion transformer is wild!

Test-Time Training (TTT) is available in video generation now! We can directly generate complete one-minute video, with great temporal and spatial coherence. We created more episodes of Tom and Jerry (my favorite cartoon in childhood) with our model. test-time-training.github.io/video-dit/

#RSS2025 NaVILA constitutes a successful attempt for VILA to drive real world robotic dogs and humanoid! Fully deployable. Money saving. Fast inference. Check out our project page: navila-bot.github.io Many more amazing things to come!

Excited to be at #ICLR2025 in person this year! Looking forward to reconnecting and making new friends.🤩 Come chat with us about Dynamic Gaussians Mesh at poster #97 tomorrow (4/26, 3–5:30pm). See you there!🥳 Website: liuisabella.com/DG-Mesh

talk by Nicklas Hansen at the world models workshop starting now!

Takeaway: Structured, reflective reasoning can be taught — even in perception. We show that generating better data can unlock stronger visual reasoning. 🌐Website: andrewliao11.github.io/LongPerceptual… 🤗Dataset: huggingface.co/datasets/andre… 📜Paper: arxiv.org/abs/2504.15362

🚨CVPR Workshop on Foundation Models + Embodied Agents Extending non-archival submission deadline to be after NeurIPS, May 17th! 🌐Website: …models-meet-embodied-agents.github.io/cvpr2025/ 📜OpenReview: openreview.net/group?id=thecv… 👥Program committee sign up form forms.gle/bL17vmr7ZbybxE… ✉️mailing list:

Supervised learning has held 3D Vision back for too long. Meet RayZer — a self-supervised 3D model trained with zero 3D labels: ❌ No supervision of camera & geometry ✅ Just RGB images And the wild part? RayZer outperforms supervised methods (as 3D labels from COLMAP is noisy)

Meet 𝐀𝐌𝐎 — our universal whole‑body controller that unleashes the 𝐟𝐮𝐥𝐥  kinematic workspace of humanoid robots to the physical world. AMO is a single policy trained with RL + Hybrid Mocap & Trajectory‑Opt. Accepted to #RSS2025. Try our open models & more 👉

Humanoids on campus! Check out our new work on context-aware locomotion

Full episode dropping soon! Geeking out with Roger Qiu on Humanoid Policy ~ Human Policy human-as-robot.github.io Co-hosted by Chris Paxton & Michael Cho - Rbt/Acc

For years, I’ve been tuning parameters for robot designs and controllers on specific tasks. Now we can automate this on dataset-scale. Introducing Co-Design of Soft Gripper with Neural Physics - a soft gripper trained in simulation to deform while handling load.

We have been focusing on policy learning for robotics for a while. But can hardware be learned as well? Check out yisha ‘s recent co-design work that learns what a soft gripper should be if we want to do better manipulation.