正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

We will make a presentation of "Developing Vision-Language-Action Model from Egocentric Videos" at TuI1I.167, 9:00-10:30 6/2, #ICRA2026 We pretrain our robot policy from solely egocentric videos, and it works with ALOHA Stationary by lab-scale teleoperation-based finetuning!

Shuhei Kurita

1,162 subscribers

10,492 次观看 • 4 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

1 条评论

Shuhei Kurita 的头像

Shuhei Kurita4 个月前

Project page: Paper: Code:

相关视频

What happens when robot world models learn from human experience at scale? 🤔 DreamDojo from NVIDIA Research is a generalist robot world model pretrained on 44K hours of egocentric human videos and then post-trained on robot data to generalize across new objects and environments. After distillation, it runs at 10 FPS for live teleoperation, policy evaluation, and model-based planning. Read the ICML paper to learn more 📄

What happens when robot world models learn from human experience at scale? 🤔 DreamDojo from NVIDIA Research is a generalist robot world model pretrained on 44K hours of egocentric human videos and then post-trained on robot data to generalize across new objects and environments. After distillation, it runs at 10 FPS for live teleoperation, policy evaluation, and model-based planning. Read the ICML paper to learn more 📄

NVIDIA Robotics

23,392 次观看 • 3 个月前

The problem with humanoid teleoperation is that it is expensive and difficult to scale Enter NVIDIA's EgoScale: - A VLA model pretrained on thousands hours of egocentric human videos. - Mid-trained via 50 hours of human + 4 hours of robot "play" data for human-robot alignment. - Fine-tuned with very few examples of task-specific robot teleoperation (100 or fewer per task). - Successfully transfers across 5-finger (Sharpa) and 3-finger (Unitree G1) robot hands. - Performance scales predictably as data increases.

The problem with humanoid teleoperation is that it is expensive and difficult to scale Enter NVIDIA's EgoScale: - A VLA model pretrained on thousands hours of egocentric human videos. - Mid-trained via 50 hours of human + 4 hours of robot "play" data for human-robot alignment. - Fine-tuned with very few examples of task-specific robot teleoperation (100 or fewer per task). - Successfully transfers across 5-finger (Sharpa) and 3-finger (Unitree G1) robot hands. - Performance scales predictably as data increases.

The Humanoid Hub

44,960 次观看 • 7 个月前

Tomorrow (Thursday morning), 9:00–10:30 AM, I'll be at Poster 326 to present our work on a new imitation learning framework, MIMIC, for training a sidewalk autopilot. Stop by if you're attending #ICRA2026 in Vienna: MIMIC (Multi-scale IMItation with Corrective expansions) trains sidewalk autopilots from teleoperation data by expanding corrective behaviors and visual diversity through generative augmentation. Project page:

Tomorrow (Thursday morning), 9:00–10:30 AM, I'll be at Poster 326 to present our work on a new imitation learning framework, MIMIC, for training a sidewalk autopilot. Stop by if you're attending #ICRA2026 in Vienna: MIMIC (Multi-scale IMItation with Corrective expansions) trains sidewalk autopilots from teleoperation data by expanding corrective behaviors and visual diversity through generative augmentation. Project page:

Bolei Zhou

11,598 次观看 • 4 个月前

Introducing Open-TeleVision: with Fully Autonomous policy video👇. We can conduct a long-horizon task with inserting 12 cans nonstop without any interruptions. We offer: 🤖 Highly precise and smooth bimanual manipulation. 📺 Active egocentric vision (with a moving neck) feedback. It is achieved by imitation learning from teleoperation: We propose a VR-based REAL-TIME teleoperation that streams the stereo video observation from the robot camera to the VR device. The robot neck moves as the human head moves, the robot hands move as the human hands move, offering the operator an intuitive experience as the human herself becomes the robot. The devils are all in the details, and how to implement things right: ✅How to perform IK/retargeting for smooth and precise control. ✅How to do all these and also stream stereo video without no latency, all in real time. We released our code here: Active head hardware design: 1/n

Introducing Open-TeleVision: with Fully Autonomous policy video👇. We can conduct a long-horizon task with inserting 12 cans nonstop without any interruptions. We offer: 🤖 Highly precise and smooth bimanual manipulation. 📺 Active egocentric vision (with a moving neck) feedback. It is achieved by imitation learning from teleoperation: We propose a VR-based REAL-TIME teleoperation that streams the stereo video observation from the robot camera to the VR device. The robot neck moves as the human head moves, the robot hands move as the human hands move, offering the operator an intuitive experience as the human herself becomes the robot. The devils are all in the details, and how to implement things right: ✅How to perform IK/retargeting for smooth and precise control. ✅How to do all these and also stream stereo video without no latency, all in real time. We released our code here: Active head hardware design: 1/n

Xiaolong Wang

25,744 次观看 • 2 年前

Egocentric Videos for Robot Training 🧵 Researchers at NYU and UC Berkeley published research where they developed a system called EgoZero that trains robots using human demonstration videos recorded with smart glasses. 👓 The system converts first-person human actions into 3D point-based state-action representations. The policies executed on a gripper-equipped robot, achieved a 70% zero-shot success rate across seven manipulation tasks, with only 20 minutes of human data per task. EgoZero stands as one of the solid empirical proofs that egocentric smart-glass video collected from everyday human behavior can serve as powerful, scalable training data for real robot learning. Vincent Liu Ademi Adeniji

Egocentric Videos for Robot Training 🧵 Researchers at NYU and UC Berkeley published research where they developed a system called EgoZero that trains robots using human demonstration videos recorded with smart glasses. 👓 The system converts first-person human actions into 3D point-based state-action representations. The policies executed on a gripper-equipped robot, achieved a 70% zero-shot success rate across seven manipulation tasks, with only 20 minutes of human data per task. EgoZero stands as one of the solid empirical proofs that egocentric smart-glass video collected from everyday human behavior can serve as powerful, scalable training data for real robot learning. Vincent Liu Ademi Adeniji

VaderResearch

23,055 次观看 • 1 年前

1/ 🧠Humans are the best robot data source! 2/ 👓Human egocentric video is rich in quantity, but poor in quality. 3/ Beyond scaling data, smarter representation and architecture matter just as much. 4/ Want an open-source framework to train your own learn-from-human-data robot policy? 🚀We introduce HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Robot-Data-Free ✦ Just 30 min of data per task ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Easy-to-Implement Let's squeeze every bit of signal out of human data! 🌐 Website: 📄 Paper: 💻 Code: 📹 Video: 🧵 1/n

1/ 🧠Humans are the best robot data source! 2/ 👓Human egocentric video is rich in quantity, but poor in quality. 3/ Beyond scaling data, smarter representation and architecture matter just as much. 4/ Want an open-source framework to train your own learn-from-human-data robot policy? 🚀We introduce HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos⬇️ ✦ Zero-Shot Human-to-Robot Transfer ✦ Robot-Data-Free ✦ Just 30 min of data per task ✦ Collect by Anyone, Anytime, Anywhere ✦ Deploy on Any Robot, Any Camera, Any Environment ✦ Open-Source & Easy-to-Implement Let's squeeze every bit of signal out of human data! 🌐 Website: 📄 Paper: 💻 Code: 📹 Video: 🧵 1/n

Zhi (Leo) Wang

127,738 次观看 • 4 个月前

Excited to announce GR00T N1, the world’s first open foundation model for humanoid robots! We are on a mission to democratize Physical AI. The power of general robot brain, in the palm of your hand - with only 2B parameters, N1 learns from the most diverse physical action dataset ever compiled and punches above its weight: - Real humanoid teleoperation data. - Large-scale simulation data: we are open-sourcing 300K+ trajectories! - Neural trajectories: we apply SOTA video generation models to “hallucinate” new synthetic data that features accurate physics in pixels. Using Jensen’s words, “systematically infinite data”! - Latent actions: we develop novel algorithms to extract action tokens from in-the-wild human videos and neural generated videos. GR00T N1 is a single end-to-end neural net, from photons to actions: - Vision-Language Model (System 2) that interprets the physical world through vision and language instructions, enabling robots to reason about their environment and instructions, and plan the right actions. - Diffusion Transformer (System 1) that “renders” smooth and precise motor actions at 120 Hz, executing the latent plan made by System 2. We deploy N1 on GR1 robot, 1X Neo robot, and a large collection of simulation benchmarks. N1 achieves up to +30% boost in diverse manipulation tasks for household and industrial settings. While humanoid robots are the main focus of N1, our model also supports cross-embodiment. We finetune it to work on the $110 HuggingFace LeRobot SO100 robot arm! Open robot brain runs on open hardware. Sounds just right. Let’s solve robotics, together, one token at a time. Links to our Whitepaper, Github repo, HuggingFace model, and open dataset page in the thread: 🧵

Excited to announce GR00T N1, the world’s first open foundation model for humanoid robots! We are on a mission to democratize Physical AI. The power of general robot brain, in the palm of your hand - with only 2B parameters, N1 learns from the most diverse physical action dataset ever compiled and punches above its weight: - Real humanoid teleoperation data. - Large-scale simulation data: we are open-sourcing 300K+ trajectories! - Neural trajectories: we apply SOTA video generation models to “hallucinate” new synthetic data that features accurate physics in pixels. Using Jensen’s words, “systematically infinite data”! - Latent actions: we develop novel algorithms to extract action tokens from in-the-wild human videos and neural generated videos. GR00T N1 is a single end-to-end neural net, from photons to actions: - Vision-Language Model (System 2) that interprets the physical world through vision and language instructions, enabling robots to reason about their environment and instructions, and plan the right actions. - Diffusion Transformer (System 1) that “renders” smooth and precise motor actions at 120 Hz, executing the latent plan made by System 2. We deploy N1 on GR1 robot, 1X Neo robot, and a large collection of simulation benchmarks. N1 achieves up to +30% boost in diverse manipulation tasks for household and industrial settings. While humanoid robots are the main focus of N1, our model also supports cross-embodiment. We finetune it to work on the $110 HuggingFace LeRobot SO100 robot arm! Open robot brain runs on open hardware. Sounds just right. Let’s solve robotics, together, one token at a time. Links to our Whitepaper, Github repo, HuggingFace model, and open dataset page in the thread: 🧵

Jim Fan

467,630 次观看 • 1 年前

One of my favorite works in the "robot learning from human videos" space that I have seen lately. This is because it offers a really coherent and unique way to represent robot skills without a lot of the downsides we see from estimation-based approaches.

One of my favorite works in the "robot learning from human videos" space that I have seen lately. This is because it offers a really coherent and unique way to represent robot skills without a lot of the downsides we see from estimation-based approaches.

Chris Paxton

78,419 次观看 • 11 个月前

What would a World Model look like if we start from a real embodied agent acting in the real world? It has to have: 1) A real, physically grounded and complex action space—not just abstract control signals. 2) Diverse, real-life scenarios and activities. Or in short: It has to be annoyingly complex—in both the action and vision space—to even get close to real life. We did an initial attempt: Whole-Body Conditioned Egocentric Video Prediction. In collaboration with @dans_t123 , Amir Bar, Yann LeCun , trevordarrell and Jitendra MALIK. (For more details, check: What we did is very simple: Predict Egocentric Video from human Actions (PEVA) - Given the past video and a future action represented by relative 3D body pose, PEVA predicts how the world looks next—from the first-person view. By conditioning on kinematic pose trajectories, structured by the joint hierarchy of the body, it learns how physical actions shape perception.

What would a World Model look like if we start from a real embodied agent acting in the real world? It has to have: 1) A real, physically grounded and complex action space—not just abstract control signals. 2) Diverse, real-life scenarios and activities. Or in short: It has to be annoyingly complex—in both the action and vision space—to even get close to real life. We did an initial attempt: Whole-Body Conditioned Egocentric Video Prediction. In collaboration with @dans_t123 , Amir Bar, Yann LeCun , trevordarrell and Jitendra MALIK. (For more details, check: What we did is very simple: Predict Egocentric Video from human Actions (PEVA) - Given the past video and a future action represented by relative 3D body pose, PEVA predicts how the world looks next—from the first-person view. By conditioning on kinematic pose trajectories, structured by the joint hierarchy of the body, it learns how physical actions shape perception.

Yutong Bai

179,812 次观看 • 1 年前

What could robots learn from the videos we already have? Join the Video to Data (V2D) Challenge to help turn human demonstrations into robot skills for 3 tracks: 🔹 4D Reconstruction 🔹 Robotic Grounding 🔹 Egocentric Video-to-Policy. You'll have the opportunity to: 🏆 Earn recognition at #CoRL2026, where the winning solutions will be announced 🛠️ Receive evaluation at scale, with top teams invited to integrate their code into a maintained repository 📋 Contribute to the challenge’s technical report Register here by September 21, 2026:

What could robots learn from the videos we already have? Join the Video to Data (V2D) Challenge to help turn human demonstrations into robot skills for 3 tracks: 🔹 4D Reconstruction 🔹 Robotic Grounding 🔹 Egocentric Video-to-Policy. You'll have the opportunity to: 🏆 Earn recognition at #CoRL2026, where the winning solutions will be announced 🛠️ Receive evaluation at scale, with top teams invited to integrate their code into a maintained repository 📋 Contribute to the challenge’s technical report Register here by September 21, 2026:

NVIDIA Robotics

13,339 次观看 • 22 天前

How well do egocentric human videos scale? Apparently, really well. Dyna Robotics has introduced Dyna-2, a world-action model (WAM) pre-trained on more than 1 million hours of real egocentric human video. It learns by jointly predicting the next video frames and the actions, and that video prediction (world modeling) is what carries learning from humans to robots. • More human video predictably improves action prediction. • Task accuracy improves across embodiments (bimanual arms and a humanoid) with zero robot data in pre-training. • Post-trained on just a few hours of robot data, mean task performance rose 20% to 53% as human pre-training scaled 1k to 1M hours. WAM vs VLA, matched apple-to-apple: the WAM hit 1.55x the VLA's success rate and won 65% of head-to-heads. Deployed zero-shot at unseen customer sites, it passed 87% vs the VLA's 46%.

How well do egocentric human videos scale? Apparently, really well. Dyna Robotics has introduced Dyna-2, a world-action model (WAM) pre-trained on more than 1 million hours of real egocentric human video. It learns by jointly predicting the next video frames and the actions, and that video prediction (world modeling) is what carries learning from humans to robots. • More human video predictably improves action prediction. • Task accuracy improves across embodiments (bimanual arms and a humanoid) with zero robot data in pre-training. • Post-trained on just a few hours of robot data, mean task performance rose 20% to 53% as human pre-training scaled 1k to 1M hours. WAM vs VLA, matched apple-to-apple: the WAM hit 1.55x the VLA's success rate and won 65% of head-to-heads. Deployed zero-shot at unseen customer sites, it passed 87% vs the VLA's 46%.

The Humanoid Hub

15,970 次观看 • 1 个月前

An interactive world model developed by NVIDIA in collaboration with academic partners. - DreamDojo turns egocentric human video data into physical intelligence. - Human data is more scalable than robotics data but lacks action labels. - To solve this, a dedicated action model extracts latent actions by identifying physics and motion deltas between frames. Training - A massive 44k hours of video data are used for pre-training. - Post-training on small-scale robot datasets maps human physics to specific robot embodiments. - An additional distillation stage converts the model into an autoregressive, few-step diffusion model, enabling real-time, action-controllable simulation. Primary Use Cases - Live Teleoperation: Controlling a robot inside a world simulation in real-time. - Model-based Planning: Previewing and curating the best actions for improved success. - Policy Evaluation: Testing robot policies in realistic, out-of-distribution scenarios. Everything that's open-sourced: weights, code, post-training dataset, eval set, and details to reproduce.

An interactive world model developed by NVIDIA in collaboration with academic partners. - DreamDojo turns egocentric human video data into physical intelligence. - Human data is more scalable than robotics data but lacks action labels. - To solve this, a dedicated action model extracts latent actions by identifying physics and motion deltas between frames. Training - A massive 44k hours of video data are used for pre-training. - Post-training on small-scale robot datasets maps human physics to specific robot embodiments. - An additional distillation stage converts the model into an autoregressive, few-step diffusion model, enabling real-time, action-controllable simulation. Primary Use Cases - Live Teleoperation: Controlling a robot inside a world simulation in real-time. - Model-based Planning: Previewing and curating the best actions for improved success. - Policy Evaluation: Testing robot policies in realistic, out-of-distribution scenarios. Everything that's open-sourced: weights, code, post-training dataset, eval set, and details to reproduce.

The Humanoid Hub

11,575 次观看 • 7 个月前

The first open-source unified world model for scalable robot manipulation: 5B-parameter open-source unified video-action world model that combines policy and world modeling to generate robot actions, predict future visuals, and evaluate task progress from observations, language, and state. The model is trained on 27.3K hours of heterogeneous data (17.8K real-robot teleop, 6.5K UMI demos, 3K egocentric human videos), enabling it to perform complex manipulation tasks like faucet connecting, bag packing, and toolbox storing as shown in demo videos. The approach supports test-time action refinement and points toward deployment-driven continuous improvement via fleet data. Thanks for sharing, Jianlan Luo (Jianlan Luo)! 📌 Resource links for τ0-WM: • Project page: • GitHub (code): • Hugging Face (model weights): • Paper (PDF): ——- Weekly robotics and AI insights. Subscribe free:

The first open-source unified world model for scalable robot manipulation: 5B-parameter open-source unified video-action world model that combines policy and world modeling to generate robot actions, predict future visuals, and evaluate task progress from observations, language, and state. The model is trained on 27.3K hours of heterogeneous data (17.8K real-robot teleop, 6.5K UMI demos, 3K egocentric human videos), enabling it to perform complex manipulation tasks like faucet connecting, bag packing, and toolbox storing as shown in demo videos. The approach supports test-time action refinement and points toward deployment-driven continuous improvement via fleet data. Thanks for sharing, Jianlan Luo (Jianlan Luo)! 📌 Resource links for τ0-WM: • Project page: • GitHub (code): • Hugging Face (model weights): • Paper (PDF): ——- Weekly robotics and AI insights. Subscribe free:

Ilir Aliu

28,276 次观看 • 4 个月前

We discovered an emergent property of VLAs like π0/π0.5/π0.6: as we scale up pre-training, the model learns to align human videos and robot data! This gives us a simple way to leverage human videos. Once π0.5 knows how to control robots, it can naturally learn from human video.

We discovered an emergent property of VLAs like π0/π0.5/π0.6: as we scale up pre-training, the model learns to align human videos and robot data! This gives us a simple way to leverage human videos. Once π0.5 knows how to control robots, it can naturally learn from human video.

Physical Intelligence

1,188,037 次观看 • 9 个月前

We present EgoReAct: Real-time 3D human reaction generation from streaming egocentric video. 🌟Reacting to streaming egocentric video is something humans do every day. We hope EgoReAct makes human motion more human-like. 🔎 What we found: existing ego-reaction data can be spatially inconsistent (e.g., moving reactions paired with fixed-camera videos), which breaks 3D grounding. 📷 What we built: HRD, a spatially aligned egocentric video–reaction dataset (3,500 pairs, 32 categories), plus a spatially aligned ViMo fix for fair evaluation. (Instead of collecting expensive ground-truth motion, we employ VDM to generate the egocentric videos.) 👁️⚡🏃 Our simple yet effective pipeline: motion tokenization for compact discrete codes + an autoregressive Transformer for online, strictly-causal generation. Metric depth and head dynamics further improve 3D spatial consistency. Project Page: ArXiv: #HumanMotion #EgocentricVision #3D #ARVR #Animation #AIGC #DeepLearning #GenerativeAI #Graphics #ComputerVision #Motion

We present EgoReAct: Real-time 3D human reaction generation from streaming egocentric video. 🌟Reacting to streaming egocentric video is something humans do every day. We hope EgoReAct makes human motion more human-like. 🔎 What we found: existing ego-reaction data can be spatially inconsistent (e.g., moving reactions paired with fixed-camera videos), which breaks 3D grounding. 📷 What we built: HRD, a spatially aligned egocentric video–reaction dataset (3,500 pairs, 32 categories), plus a spatially aligned ViMo fix for fair evaluation. (Instead of collecting expensive ground-truth motion, we employ VDM to generate the egocentric videos.) 👁️⚡🏃 Our simple yet effective pipeline: motion tokenization for compact discrete codes + an autoregressive Transformer for online, strictly-causal generation. Metric depth and head dynamics further improve 3D spatial consistency. Project Page: ArXiv: #HumanMotion #EgocentricVision #3D #ARVR #Animation #AIGC #DeepLearning #GenerativeAI #Graphics #ComputerVision #Motion

Zhiyang (Frank) Dou

11,413 次观看 • 9 个月前

Chinese robotics company Astribot released their latest World-Action Model (WAM), Lumo-2. Technical breakdown: - based on a frozen 🥶 Qwen-3.5 4B VLM - trained in 3 progressive stages: 1. Action is aligned with latent world dynamics (an abstract representation of action). Real-world actions are anchored to physical constraints, while the latent space is guided to focus on motion-relevant changes. This bidirectional relationship makes the model physically grounded -> critical for a world model. 2. Action is aligned with vision and language. Reusing the vision backbone and action encoder from the frozen VLM, the authors add a custom vocabulary (for new actions), a semantic module, an action decoder, and an action projector. This aligns the (new) action representations with the (existing) vision-language semantic space. Most importantly: it builds a direct mapping from natural-language instructions to motor execution. 3. End-to-end training on language, video, and robot data. Only the new modules (everything outside the frozen backbone) are trained end-to-end across temporal reasoning, physical understanding, long-horizon, and dexterous manipulation. At the end of the day, Lumo-2 is not the best on benchmarks, but that's not the point. What's genuinely new: - a way to combine latent world modeling and action generation through progressive alignment - a physically-grounded latent dynamics space - it lifts performance on unseen objects using un-annotated human egocentric video + Vision Pro captures, no special transfer algorithm needed Why it matters: - the whole model is thin trainable adapters (semantic module, action decoder/projector) on a frozen 4B backbone (cheap) - that scale is suited for real-time embedded inference (~2.71× decode speedup, no accuracy loss) - its real moat is long-horizon execution, where the added temporal memory pays off far more than on any other task As a result, this robot can now make your latte (5x sped up video):

Chinese robotics company Astribot released their latest World-Action Model (WAM), Lumo-2. Technical breakdown: - based on a frozen 🥶 Qwen-3.5 4B VLM - trained in 3 progressive stages: 1. Action is aligned with latent world dynamics (an abstract representation of action). Real-world actions are anchored to physical constraints, while the latent space is guided to focus on motion-relevant changes. This bidirectional relationship makes the model physically grounded -> critical for a world model. 2. Action is aligned with vision and language. Reusing the vision backbone and action encoder from the frozen VLM, the authors add a custom vocabulary (for new actions), a semantic module, an action decoder, and an action projector. This aligns the (new) action representations with the (existing) vision-language semantic space. Most importantly: it builds a direct mapping from natural-language instructions to motor execution. 3. End-to-end training on language, video, and robot data. Only the new modules (everything outside the frozen backbone) are trained end-to-end across temporal reasoning, physical understanding, long-horizon, and dexterous manipulation. At the end of the day, Lumo-2 is not the best on benchmarks, but that's not the point. What's genuinely new: - a way to combine latent world modeling and action generation through progressive alignment - a physically-grounded latent dynamics space - it lifts performance on unseen objects using un-annotated human egocentric video + Vision Pro captures, no special transfer algorithm needed Why it matters: - the whole model is thin trainable adapters (semantic module, action decoder/projector) on a frozen 4B backbone (cheap) - that scale is suited for real-time embedded inference (~2.71× decode speedup, no accuracy loss) - its real moat is long-horizon execution, where the added temporal memory pays off far more than on any other task As a result, this robot can now make your latte (5x sped up video):

Léo

32,513 次观看 • 2 个月前

Announcing DreamDojo: our open-source, interactive world model that takes robot motor controls and generates the future in pixels. No engine, no meshes, no hand-authored dynamics. It's Simulation 2.0. Time for robotics to take the bitter lesson pill. Real-world robot learning is bottlenecked by time, wear, safety, and resets. If we want Physical AI to move at pretraining speed, we need a simulator that adapts to pretraining scale with as little human engineering as possible. Our key insights: (1) human egocentric videos are a scalable source of first-person physics; (2) latent actions make them "robot-readable" across different hardware; (3) real-time inference unlocks live teleop, policy eval, and test-time planning *inside* a dream. We pre-train on 44K hours of human videos: cheap, abundant, and collected with zero robot-in-the-loop. Humans have already explored the combinatorics: we grasp, pour, fold, assemble, fail, retry—across cluttered scenes, shifting viewpoints, changing light, and hour-long task chains—at a scale no robot fleet could match. The missing piece: these videos have no action labels. So we introduce latent actions: a unified representation inferred directly from videos that captures "what changed between world states" without knowing the underlying hardware. This lets us train on any first-person video as if it came with motor commands attached. As a result, DreamDojo generalizes zero-shot to objects and environments never seen in any robot training set, because humans saw them first. Next, we post-train onto each robot to fit its specific hardware. Think of it as separating "how the world looks and behaves" from "how this particular robot actuates." The base model follows the general physical rules, then "snaps onto" the robot's unique mechanics. It's kind of like loading a new character and scene assets into Unreal Engine, but done through gradient descent and generalizes far beyond the post-training dataset. A world simulator is only useful if it runs fast enough to close the loop. We train a real-time version of DreamDojo that runs at 10 FPS, stable for over a minute of continuous rollout. This unlocks exciting possibilities: - Live teleoperation *inside* a dream. Connect a VR controller, stream actions into DreamDojo, and teleop a virtual robot in real time. We demo this on Unitree G1 with a PICO headset and one RTX 5090. - Policy evaluation. You can benchmark a policy checkpoint in DreamDojo instead of the real world. The simulated success rates strongly correlate with real-world results - accurate enough to rank checkpoints without burning a single motor. - Model-based planning. Sample multiple action proposals → simulate them all in parallel → pick the best future. Gains +17% real-world success out of the box on a fruit packing task. We open-source everything!! Weights, code, post-training dataset, eval set, and whitepaper with tons of details to reproduce. DreamDojo is based on NVIDIA Cosmos, which is open-weight too. 2026 is the year of World Models for physical AI. We want you to build with us. Happy scaling! Links in thread:

Announcing DreamDojo: our open-source, interactive world model that takes robot motor controls and generates the future in pixels. No engine, no meshes, no hand-authored dynamics. It's Simulation 2.0. Time for robotics to take the bitter lesson pill. Real-world robot learning is bottlenecked by time, wear, safety, and resets. If we want Physical AI to move at pretraining speed, we need a simulator that adapts to pretraining scale with as little human engineering as possible. Our key insights: (1) human egocentric videos are a scalable source of first-person physics; (2) latent actions make them "robot-readable" across different hardware; (3) real-time inference unlocks live teleop, policy eval, and test-time planning inside a dream. We pre-train on 44K hours of human videos: cheap, abundant, and collected with zero robot-in-the-loop. Humans have already explored the combinatorics: we grasp, pour, fold, assemble, fail, retry—across cluttered scenes, shifting viewpoints, changing light, and hour-long task chains—at a scale no robot fleet could match. The missing piece: these videos have no action labels. So we introduce latent actions: a unified representation inferred directly from videos that captures "what changed between world states" without knowing the underlying hardware. This lets us train on any first-person video as if it came with motor commands attached. As a result, DreamDojo generalizes zero-shot to objects and environments never seen in any robot training set, because humans saw them first. Next, we post-train onto each robot to fit its specific hardware. Think of it as separating "how the world looks and behaves" from "how this particular robot actuates." The base model follows the general physical rules, then "snaps onto" the robot's unique mechanics. It's kind of like loading a new character and scene assets into Unreal Engine, but done through gradient descent and generalizes far beyond the post-training dataset. A world simulator is only useful if it runs fast enough to close the loop. We train a real-time version of DreamDojo that runs at 10 FPS, stable for over a minute of continuous rollout. This unlocks exciting possibilities: - Live teleoperation inside a dream. Connect a VR controller, stream actions into DreamDojo, and teleop a virtual robot in real time. We demo this on Unitree G1 with a PICO headset and one RTX 5090. - Policy evaluation. You can benchmark a policy checkpoint in DreamDojo instead of the real world. The simulated success rates strongly correlate with real-world results - accurate enough to rank checkpoints without burning a single motor. - Model-based planning. Sample multiple action proposals → simulate them all in parallel → pick the best future. Gains +17% real-world success out of the box on a fruit packing task. We open-source everything!! Weights, code, post-training dataset, eval set, and whitepaper with tons of details to reproduce. DreamDojo is based on NVIDIA Cosmos, which is open-weight too. 2026 is the year of World Models for physical AI. We want you to build with us. Happy scaling! Links in thread:

Jim Fan

230,266 次观看 • 7 个月前

🤖What if a robot could perform a new task just from a natural language command, with zero demonstrations? Our new work, NovaFlow, makes it possible! We use pre-trained video generative model to create a video of the task, then translate it into a plan for real-world robot execution. 1/6 #Robotics #AI #ZeroShot #Manipulation

🤖What if a robot could perform a new task just from a natural language command, with zero demonstrations? Our new work, NovaFlow, makes it possible! We use pre-trained video generative model to create a video of the task, then translate it into a plan for real-world robot execution. 1/6 #Robotics #AI #ZeroShot #Manipulation

Hongyu Li

106,075 次观看 • 1 年前