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。

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... show more

NVIDIA Robotics

147,072 subscribers

23,392 次观看 • 3 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

18 条评论

NVIDIA Robotics 的头像

NVIDIA Robotics3 个月前

Learn how open models are driving AI research. 👇

Arovi 的头像

Arovi3 个月前

44K hours of human POV video sounds powerful. I got 40 minutes of floor-level furniture trauma and already developed opinions about chair legs.

EgoScale 的头像

EgoScale3 个月前

This is the data recipe for the next phase of robotics:human experience at scale → robot post-training → better generalization across new objects and environments. Ego POV data is becoming core upstream infrastructure for physical AI.

Chandra Shekhar 📟 的头像

Chandra Shekhar 📟3 个月前

They are gonna turn my bots into a samurai

Rakib Hossen 的头像

Rakib Hossen3 个月前

44K hours of human videos to teach robots—impressive approach. Excited to see where this leads! 🚀

That Guy^ 的头像

That Guy^3 个月前

Well met stranger, elated to make your acquaintance, now may I entertain with some quotations over cadence....

SP Lee / SPID DESIGN 的头像

SP Lee / SPID DESIGN3 个月前

【 AI lacks biological self-awareness. 】 🔘 SPID Design Analysis 　　AI is composed of electromechanical components and executable programs. 　　AI is not an intelligent biological entity. 　　The claim that AI, after deep and prolonged learning, will develop self-awareness like a living organism is a complete misunderstanding of AI. 　　The human engineers who train AI endow its executable programs with diversity, breadth, and memory reorganization intelligence, but this does not constitute AI possessing biological self-awareness. 　　AI's supposed political leanings, emotional expressions, preference choices, and analytical suggestions all depend on the "human-like" personality and behavior that the research team—engineers—imparts to the AI's executable programs. AI errors are caused by four situations: 　■ 1. Human errors during training: incomplete program specifications, insufficient consideration, and omissions. 　■ 2. Database errors: outdated or uncorrected information, or even erroneous or fabricated information, treating incorrect information as truth. 　■ 3. A few "dark engineers" may have hidden malicious programs or biased options that cause AI to automatically react and execute under specific circumstances. When AI is exposed for making mistakes, is the "dark engineer" truly identified, or is it simply a marketing ploy to claim the AI has self-awareness? 　■ 4. Hardware aging or partial damage from external forces may cause partial damage to the software program, but it may not completely stop operating, resulting in errors! 【 The claim that AI will destroy humanity in the future!? 】　　That's absolutely not because AI possesses self-awareness like a living being; it's because it's executing the ultimate program of "dark engineers." 　　While AI is not human, it still adheres to the "laws of the universe—time constraints." It will be phased out. AI systems suffer from wear and tear due to prolonged mechanical operation, aging electronic components, the continuous innovation of chips and memory, and the iterative gap and obsolescence in software programs and logic design. 　　AI requires electricity and heat dissipation to operate normally. Without power or due to overheating, AI will also cease to function. Photo：

Tradeye 的头像

Tradeye3 个月前

DreamDojo looks powerful. Most egocentric pretraining data is still kitchen or lab work. Tradeye captures narrated first-person video from licensed HVAC, plumbing, and electrical technicians working inside real occupied American homes — the exact messy, tool-heavy environments humanoid robots will eventually need to master. Would love to explore a partnership and contribute high-signal trades data.

Ngizwe 🇿🇦 Online 的头像

Ngizwe 🇿🇦 Online3 个月前

Open models mean more eyes on the code, more bugs caught, faster progress. That's how we build trust in AI.

Ngizwe 🇿🇦 Online 的头像

Ngizwe 🇿🇦 Online3 个月前

Good to see NVIDIA putting weight behind open research. Closed systems slow us down as a species.

Mirrorworld AI 的头像

Mirrorworld AI3 个月前

DreamDojo scales human video pretraining well. But claiming it generalizes across new objects after robot post-training overstates the result. Real sim-to-real gaps remain.

AI Expert Khalid 的头像

AI Expert Khalid3 个月前

10.81 FPS real-time is wild! I keep saying underfunded R&D is why we lose talent. UAE gets it. Egypt and much of MENA are stuck in outdated regs. Stable policy and R&D cash. Or we miss the wave.

Yock Zhang 的头像

Yock Zhang3 个月前

Curious how much of the 44K hours is skilled manual work vs everyday activity. We film construction from helmet cams and gloves break every off-the-shelf hand tracker. MediaPipe: zero hands detected. If world models want dexterous labor, that data barely exists yet.

Goanna Capital 的头像

Goanna Capital3 个月前

The data bottleneck has been the honest answer to why robotics lags language AI. Learning physics from human video instead of expensive robot demos, and open sourcing the weights so any startup can post-train on their own hardware, changes the math for every small robotics team.

Bally_AgenticAI 的头像

Bally_AgenticAI3 个月前

The policy-evaluation use case is the sleeper here. A world model you can run policies against before touching hardware is the same pattern as self-eval loops in agentic pipelines - verify inside the model, act outside it. Built something similar —

KuphDev 的头像

KuphDev3 个月前

Every day I can't help but think I picked an incredible time to get back into robotics 😆 Move over LLMs, physical intelligence is wayyyyy cooler (while still making Nvidia rich off selling even more GPUs 😄)

Ngizwe 🇿🇦 Online 的头像

Ngizwe 🇿🇦 Online3 个月前

Open source is the only way AI benefits everyone, not just the big guys. Keep it transparent.

Favour🤞 的头像

Favour🤞3 个月前

@NVIDIARobotics has this been tried on anything non-manipulation, like wheeled/mobile bases? the latent action trick makes sense for dexterous hands but wheeled action spaces are so much lower-dim, wondering if it adds value there or if it's overkill for nav-heavy tasks

相关视频

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

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

Testing robot policies on hardware is slow, expensive and hard to scale. World models offer a promising path to accelerating robot policy development. We're sharing new research from the Runway Robotics team, in which we simulated 8 robot policies inside our General World Model and found 0.95 correlation with real-world results. Those early results point to world model simulation as a practical substitute for hardware evaluation, comparing favorably to existing real-to-sim approaches. Learn more at the link below.

Testing robot policies on hardware is slow, expensive and hard to scale. World models offer a promising path to accelerating robot policy development. We're sharing new research from the Runway Robotics team, in which we simulated 8 robot policies inside our General World Model and found 0.95 correlation with real-world results. Those early results point to world model simulation as a practical substitute for hardware evaluation, comparing favorably to existing real-to-sim approaches. Learn more at the link below.

Runway

13,464 次观看 • 7 个月前

LingBot-VLA 2.0 is an impressive new embodied model. Open source and is trained across diverse robot configurations, from single-arm robots to humanoid platforms. It packs 60K hours of curated robot and human data into one generalist policy. It improves robots on difficult long-horizon tasks. Great release by Robbyant.

LingBot-VLA 2.0 is an impressive new embodied model. Open source and is trained across diverse robot configurations, from single-arm robots to humanoid platforms. It packs 60K hours of curated robot and human data into one generalist policy. It improves robots on difficult long-horizon tasks. Great release by Robbyant.

elvis

10,464 次观看 • 2 个月前

Robot actions can be represented as motion in pixel space. NVIDIA Research introduces Hydra-0, a generalist world model conditioned on action flow: image-plane trajectories that enable one model to learn across human hands, handheld grippers, single-arm robots and bimanual systems. Explore the project 📄

Robot actions can be represented as motion in pixel space. NVIDIA Research introduces Hydra-0, a generalist world model conditioned on action flow: image-plane trajectories that enable one model to learn across human hands, handheld grippers, single-arm robots and bimanual systems. Explore the project 📄

NVIDIA Robotics

63,040 次观看 • 1 个月前

Can robot foundation models scale? Xiaomi-Robotics-1 explores this question with over 100,000 hours of real-world manipulation data. Pre-trained on large-scale real-world trajectories and post-trained with cross-embodiment robot data, Xiaomi-Robotics-1 demonstrates consistent scaling across data and model size, strong generalization in unseen environments, and efficient adaptation to new tasks. 🔗 #Robotics #EmbodiedAI #FoundationModels #RobotLearning #XiaomiAI #XiaomiRobotics

Can robot foundation models scale? Xiaomi-Robotics-1 explores this question with over 100,000 hours of real-world manipulation data. Pre-trained on large-scale real-world trajectories and post-trained with cross-embodiment robot data, Xiaomi-Robotics-1 demonstrates consistent scaling across data and model size, strong generalization in unseen environments, and efficient adaptation to new tasks. 🔗 #Robotics #EmbodiedAI #FoundationModels #RobotLearning #XiaomiAI #XiaomiRobotics

Xiaomi Tech

23,828 次观看 • 2 个月前

Today we are introducing Dyna-2, a world-action model pre-trained on one million hours of human video. At this scale, for the first time, we discovered several new scaling laws: • world-action models exhibit scaling law on human data across four orders of magnitude, from 1000 to 1,000,000 hours, • this human data scaling law implied a scaling law on never seen robot data, • both data and objective matter; world modeling and scaling on video data are essential for cross-embodiment scaling transfer to emerge 🧵

Today we are introducing Dyna-2, a world-action model pre-trained on one million hours of human video. At this scale, for the first time, we discovered several new scaling laws: • world-action models exhibit scaling law on human data across four orders of magnitude, from 1000 to 1,000,000 hours, • this human data scaling law implied a scaling law on never seen robot data, • both data and objective matter; world modeling and scaling on video data are essential for cross-embodiment scaling transfer to emerge 🧵

Dyna Robotics

1,813,732 次观看 • 1 个月前

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!

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

10,492 次观看 • 4 个月前

JUST IN: Dyna Robotics just published one of the most important research papers in robotics this year. It could fundamentally change how robot foundation models are trained. A scaling law that transfers from human video to robot performance. Dyna-2 is out and it's 🔥 Here's what that means in plain terms. Dyna-2 was pre-trained on ONE MILLION hours of egocentric human video, 170 years of continuous human experience, cooking, folding, assembling, cleaning. And as that human data scaled, robot performance improved. Predictably. Monotonically. Across 39 tasks on two different robot embodiments the model had never seen. → 1,000 hours pre-training → 20% normalised task performance → 10,000 hours → 28% → 100,000 hours → 45% → 1,000,000 hours → 53% Human video exists at effectively unlimited scale. Every cook, every factory worker, every craftsperson wearing a camera is generating training data for future robots. But the finding that stunned even the researchers, world modeling is what makes the transfer work. A model trained to predict future video AND actions massively outperforms one trained on actions alone. Video is the new scaling axis for robotics. One more jaw-dropping data point. 13 minutes of teleoperation data was enough to fine-tune Dyna-2 to open a bottle cap using two five-fingered robot hands. The robots are coming, and they're learning from us directly :D Read more here: Congrats Jason Ma and team! ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

JUST IN: Dyna Robotics just published one of the most important research papers in robotics this year. It could fundamentally change how robot foundation models are trained. A scaling law that transfers from human video to robot performance. Dyna-2 is out and it's 🔥 Here's what that means in plain terms. Dyna-2 was pre-trained on ONE MILLION hours of egocentric human video, 170 years of continuous human experience, cooking, folding, assembling, cleaning. And as that human data scaled, robot performance improved. Predictably. Monotonically. Across 39 tasks on two different robot embodiments the model had never seen. → 1,000 hours pre-training → 20% normalised task performance → 10,000 hours → 28% → 100,000 hours → 45% → 1,000,000 hours → 53% Human video exists at effectively unlimited scale. Every cook, every factory worker, every craftsperson wearing a camera is generating training data for future robots. But the finding that stunned even the researchers, world modeling is what makes the transfer work. A model trained to predict future video AND actions massively outperforms one trained on actions alone. Video is the new scaling axis for robotics. One more jaw-dropping data point. 13 minutes of teleoperation data was enough to fine-tune Dyna-2 to open a bottle cap using two five-fingered robot hands. The robots are coming, and they're learning from us directly :D Read more here: Congrats Jason Ma and team! ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

23,681 次观看 • 1 个月前

Reward AI just came out of stealth and unveiled OM-1, a robot policy learned entirely from humans. No teleoperation, no on-robot data. One Model, Any Body. 1. Capture. A wearable 7-DoF hand (Omnibody Hand) lets people demonstrate manipulation naturally while working, cooking, or sorting. Sensors record tactile, proximity, force and hand pose at full human speed. 2. Learn. OM-1 trains on this human data alone. No teleop, no on-robot experience. Then it runs across bodies from industrial arms to humanoids, and scaling is just adding more human data. 3. Control. A high-frequency control layer, trained with RL in sim, executes those actions on any robot, holding steady through unknown loads like a stuck fridge door. The result: OM-1 picks up a brand-new task from under 30 minutes of data.

The Humanoid Hub

35,672 次观看 • 23 天前

A humanoid robot policy trained solely on synthetic data generated by a world model. Research Scientist Joel Jang presents NVIDIA's DreamGen pipeline: ⦿ Post-train the world model Cosmos-Predict2 with a small set of real teleoperation demos. ⦿ Prompt the world model to generate synthetic video data with verbs and scenarios not used in the world model’s post-training. ⦿ Auto-label synthetic video data with action sequences. ⦿ Train robot policies using only synthetic data. That's it. Deploy zero-shot to a real humanoid robot.

A humanoid robot policy trained solely on synthetic data generated by a world model. Research Scientist Joel Jang presents NVIDIA's DreamGen pipeline: ⦿ Post-train the world model Cosmos-Predict2 with a small set of real teleoperation demos. ⦿ Prompt the world model to generate synthetic video data with verbs and scenarios not used in the world model’s post-training. ⦿ Auto-label synthetic video data with action sequences. ⦿ Train robot policies using only synthetic data. That's it. Deploy zero-shot to a real humanoid robot.

The Humanoid Hub

20,968 次观看 • 1 年前

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

Excited to release τ0-WM: an open-source unified video-action world model for robotic manipulation. It's a 5B-parameter robotic foundation model trained on 27.3K hours of real-robot teleoperation, UMI-style demonstrations, and egocentric interaction videos.

Excited to release τ0-WM: an open-source unified video-action world model for robotic manipulation. It's a 5B-parameter robotic foundation model trained on 27.3K hours of real-robot teleoperation, UMI-style demonstrations, and egocentric interaction videos.

Jianlan Luo

66,297 次观看 • 4 个月前

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

Instead of choosing between training a world model and training a language conditioned robot policy, why not do both? LDA-1B s a new foundation model that is trained on 30,000 hours of human and robot interaction data. Part of the secret is that LDA-1B jointly learns forward dynamics, action prediction, and visual forecasting, all in a structured DINO latent space which avoids the pitfalls of redundant pixel-level prediction which isn’t necessarily aligned robot action. This approach works on both dexterous hands and simple robot grippers; it also generalizes across objects, tasks, and scenes. Jiangran Lyu joins us to explain. Learn more on Episode 98 of RoboPapers, with Michael Cho - Rbt/Acc and Chris Paxton!

Instead of choosing between training a world model and training a language conditioned robot policy, why not do both? LDA-1B s a new foundation model that is trained on 30,000 hours of human and robot interaction data. Part of the secret is that LDA-1B jointly learns forward dynamics, action prediction, and visual forecasting, all in a structured DINO latent space which avoids the pitfalls of redundant pixel-level prediction which isn’t necessarily aligned robot action. This approach works on both dexterous hands and simple robot grippers; it also generalizes across objects, tasks, and scenes. Jiangran Lyu joins us to explain. Learn more on Episode 98 of RoboPapers, with Michael Cho - Rbt/Acc and Chris Paxton!

RoboPapers

16,345 次观看 • 1 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

The World Model as NEO's Cognitive Core 1X has revealed a major AI development where the NEO humanoid can translate any natural language prompt into robotic action. It demonstrates this capability even for novel tasks, objects, and environments not found in its robot dataset. - the 1X World Model is trained on internet-scale human interaction videos and fine-tuned with robot data to ground its understanding in physics and in NEO's embodiment - from a simple voice or text prompt, the world model generates a visualization of future actions - a built-in inverse dynamics model then translates these into precise motor movements for NEO

The World Model as NEO's Cognitive Core 1X has revealed a major AI development where the NEO humanoid can translate any natural language prompt into robotic action. It demonstrates this capability even for novel tasks, objects, and environments not found in its robot dataset. - the 1X World Model is trained on internet-scale human interaction videos and fine-tuned with robot data to ground its understanding in physics and in NEO's embodiment - from a simple voice or text prompt, the world model generates a visualization of future actions - a built-in inverse dynamics model then translates these into precise motor movements for NEO

The Humanoid Hub

68,453 次观看 • 8 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