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。

Introducing Axis Dataset V1 - the simulation dataset for scalable robot manipulation. Can noisy, crowdsourced simulation data support embodied pretraining? Yes—with enough coverage and diversity. Continual pretraining on AXIS dataset V1 lifts π0.5 from 83.9% to 88.8% on LIBERO-Plus, with performance improving consistently as the pretraining data scales from... show more

Axis Robotics

57,229 subscribers

96,538 次观看 • 1 个月前 •via X (Twitter)

科学技术教育

Anya Rossi• Live Now

Private livecam show

34 条评论

Stone Tao 的头像

Stone Tao1 个月前

don’t want to put too much shade on the progress (i think sim datasets can be good ideas if executed well!), but improving on libero (which is also a sim benchmark) means very little unfortunately

Thor 的头像

Thor1 个月前

Let's goo 🔥

Hazzy 的头像

Hazzy1 个月前

you edit too much intern

KABILA 的头像

KABILA1 个月前

Cookingg

Jontu 的头像

Jontu1 个月前

This is a strong proof that diverse simulation data can meaningfully improve embodied AI.

boyBrands_ 的头像

boyBrands_1 个月前

He has arrived🔥

Sami 的头像

Sami1 个月前

So much excited for this

Apurbo 的头像

Apurbo1 个月前

We are so excited about this 🤘

AL Shifat 的头像

AL Shifat1 个月前

Axis going to make history

dviet.eth 的头像

dviet.eth1 个月前

This is exactly the kind of infrastructure Physical AI needs. Quality data compounds, and so does robot intelligence. Excited to see Axis Dataset V1 keep growing 😆

Noodle 的头像

Noodle1 个月前

cool to see those teleop sessions actually pushing the libero-plus numbers up

Acee (✱,✱) 的头像

Acee (✱,✱)1 个月前

Keep cooking 👌

Piggie 🧡|Bird 的头像

Piggie 🧡|Bird1 个月前

Overall success rate increased by 5.8%.

0xmowra.eth 的头像

0xmowra.eth1 个月前

Axis V1 dataset boosting robot learning with crowdsourced simulation data impressive

Red 的头像

Red1 个月前

Great work

| MinhDucc | Shark 的头像

| MinhDucc | Shark1 个月前

Axis LFG 😍

sasafonda✱,✱) 的头像

sasafonda✱,✱)1 个月前

gaxis

namnew 的头像

namnew1 个月前

gAxis

VINAY (✱,✱) 的头像

VINAY (✱,✱)1 个月前

Axis cooking 🔥

savagezz 的头像

savagezz1 个月前

cooking something

Hoangdong (✱,✱) 的头像

Hoangdong (✱,✱)1 个月前

gAxis

clover 的头像

clover1 个月前

move gaxis

vinhle (✱,✱) 🍚 ⛓ ./ 的头像

vinhle (✱,✱) 🍚 ⛓ ./1 个月前

Exciting to see the potential of crowdsourced simulation data in robotics, and impressive results from continual pretraining on Axis Dataset V1.

Aurora Lux 的头像

Aurora Lux1 个月前

AXIS Dataset V1 is a strong step forward for embodied AI—scalable, diverse simulation data is proving to be a powerful foundation for better robot learning. 🤖🚀

cbeta.base.eth 的头像

cbeta.base.eth1 个月前

lets goo team fight fight fight

rara 的头像

rara1 个月前

gaxis lfg

Linh (✱,✱) 的头像

Linh (✱,✱)1 个月前

Axis Dataset V1 is impressive

Nova 的头像

Nova1 个月前

The bot is terrible. Please update it. Its performance is like a primitive human living in a cave 😤

MAS_Enpici (✱,✱) 的头像

MAS_Enpici (✱,✱)1 个月前

big update and new insight

yellowopus 的头像

yellowopus1 个月前

lets cooking bro

KADAFER 的头像

KADAFER1 个月前

let him cook 🔥

dondonat (✱,✱) 的头像

dondonat (✱,✱)1 个月前

gaxis keep building

sagar 的头像

sagar1 个月前

Let's goo 🙌

Gujil Ruipa 的头像

Gujil Ruipa1 个月前

Open datasets like this can accelerate robotics research through collaborative innovation

相关视频

This giant free dataset could make helper robots way smarter, way faster: An open-source robotics stack from Berkeley AI researchers featuring the largest teleoperation dataset released to date with over 3,500 hours of bimanual manipulation data across 200 tasks. The video showcases autonomous bimanual robot performance on dexterous tasks including box folding, Lego sorting, AirPod insertion, t-shirt folding, backpack packing, and box unlocking using learned policies. Sim-to-real correlations, training insights like flow loss predicting real-world success, lightweight infrastructure for DAgger interventions Thank you for sharing, Ritvik Singh, and everyone else who contributed to this! Links to the paper plus dataset at under permissive licensing. ——- Weekly robotics and AI insights. Subscribe free:

This giant free dataset could make helper robots way smarter, way faster: An open-source robotics stack from Berkeley AI researchers featuring the largest teleoperation dataset released to date with over 3,500 hours of bimanual manipulation data across 200 tasks. The video showcases autonomous bimanual robot performance on dexterous tasks including box folding, Lego sorting, AirPod insertion, t-shirt folding, backpack packing, and box unlocking using learned policies. Sim-to-real correlations, training insights like flow loss predicting real-world success, lightweight infrastructure for DAgger interventions Thank you for sharing, Ritvik Singh, and everyone else who contributed to this! Links to the paper plus dataset at under permissive licensing. ——- Weekly robotics and AI insights. Subscribe free:

Ilir Aliu

13,264 次观看 • 3 个月前

In my past research experience, finding or developing an appropriate simulation environment, dataset, and benchmark has always been a challenge. Missing features, limited support, or unexpected bugs often occupied my days and nights. Moreover, current simulation platforms are relatively fragmented—making it challenging to replicate the success of the RT-X dataset in unifying community efforts. Introducing RoboVerse, we provide a unified platform, dataset, and benchmark for scalable and generalizable robot learning. We hope to build a shared foundation to combine the community efforts. RoboVerse includes: MetaSim: We carefully designed a configuration system and a universal interface to align current robotic simulators. With MetaSim, you can use any simulator with the same code—bringing together the community’s diverse efforts under one framework! RoboVerse Dataset and Benchmark: We unify popular simulation environments and benchmarks into a single cohesive system and introduce the RoboVerse dataset—a large-scale, high-quality synthetic dataset. Additionally, we propose a standardized benchmark across both imitation learning and reinforcement learning. A cool feature enabled by our unified framework: Hybrid Simulation! You can now integrate physics engines and renderers from different simulators—e.g., using MuJoCo precise physics with Isaac photorealistic rendering. This not only elevates simulation fidelity but also significantly enhances real-world transfer performance across complex robotic applications. Hopefully, our team’s efforts could serve the robotic community to thrive vibrantly in the years to come. RoboVerse is open-sourced🥳!!! Project Page: Documentation: Github Repo: Paper:

In my past research experience, finding or developing an appropriate simulation environment, dataset, and benchmark has always been a challenge. Missing features, limited support, or unexpected bugs often occupied my days and nights. Moreover, current simulation platforms are relatively fragmented—making it challenging to replicate the success of the RT-X dataset in unifying community efforts. Introducing RoboVerse, we provide a unified platform, dataset, and benchmark for scalable and generalizable robot learning. We hope to build a shared foundation to combine the community efforts. RoboVerse includes: MetaSim: We carefully designed a configuration system and a universal interface to align current robotic simulators. With MetaSim, you can use any simulator with the same code—bringing together the community’s diverse efforts under one framework! RoboVerse Dataset and Benchmark: We unify popular simulation environments and benchmarks into a single cohesive system and introduce the RoboVerse dataset—a large-scale, high-quality synthetic dataset. Additionally, we propose a standardized benchmark across both imitation learning and reinforcement learning. A cool feature enabled by our unified framework: Hybrid Simulation! You can now integrate physics engines and renderers from different simulators—e.g., using MuJoCo precise physics with Isaac photorealistic rendering. This not only elevates simulation fidelity but also significantly enhances real-world transfer performance across complex robotic applications. Hopefully, our team’s efforts could serve the robotic community to thrive vibrantly in the years to come. RoboVerse is open-sourced🥳!!! Project Page: Documentation: Github Repo: Paper:

Haoran Geng

84,318 次观看 • 1 年前

PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking paper page: introduce PointOdyssey, a large-scale synthetic dataset, and data generation framework, for the training and evaluation of long-term fine-grained tracking algorithms. Our goal is to advance the state-of-the-art by placing emphasis on long videos with naturalistic motion. Toward the goal of naturalism, we animate deformable characters using real-world motion capture data, we build 3D scenes to match the motion capture environments, and we render camera viewpoints using trajectories mined via structure-from-motion on real videos. We create combinatorial diversity by randomizing character appearance, motion profiles, materials, lighting, 3D assets, and atmospheric effects. Our dataset currently includes 104 videos, averaging 2,000 frames long, with orders of magnitude more correspondence annotations than prior work. We show that existing methods can be trained from scratch in our dataset and outperform the published variants. Finally, we introduce modifications to the PIPs point tracking method, greatly widening its temporal receptive field, which improves its performance on PointOdyssey as well as on two real-world benchmarks.

PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking paper page: introduce PointOdyssey, a large-scale synthetic dataset, and data generation framework, for the training and evaluation of long-term fine-grained tracking algorithms. Our goal is to advance the state-of-the-art by placing emphasis on long videos with naturalistic motion. Toward the goal of naturalism, we animate deformable characters using real-world motion capture data, we build 3D scenes to match the motion capture environments, and we render camera viewpoints using trajectories mined via structure-from-motion on real videos. We create combinatorial diversity by randomizing character appearance, motion profiles, materials, lighting, 3D assets, and atmospheric effects. Our dataset currently includes 104 videos, averaging 2,000 frames long, with orders of magnitude more correspondence annotations than prior work. We show that existing methods can be trained from scratch in our dataset and outperform the published variants. Finally, we introduce modifications to the PIPs point tracking method, greatly widening its temporal receptive field, which improves its performance on PointOdyssey as well as on two real-world benchmarks.

AK

122,567 次观看 • 3 年前

I don’t know if we live in a Matrix, but I know for sure that robots will spend most of their lives in simulation. Let machines train machines. I’m excited to introduce DexMimicGen, a massive-scale synthetic data generator that enables a humanoid robot to learn complex skills from only a handful of human demonstrations. Yes, as few as 5! DexMimicGen addresses the biggest pain point in robotics: where do we get data? Unlike with LLMs, where vast amounts of texts are readily available, you cannot simply download motor control signals from the internet. So researchers teleoperate the robots to collect motion data via XR headsets. They have to repeat the same skill over and over and over again, because neural nets are data hungry. This is a very slow and uncomfortable process. At NVIDIA, we believe the majority of high-quality tokens for robot foundation models will come from simulation. What DexMimicGen does is to trade GPU compute time for human time. It takes one motion trajectory from human, and multiplies into 1000s of new trajectories. A robot brain trained on this augmented dataset will generalize far better in the real world. Think of DexMimicGen as a learning signal amplifier. It maps a small dataset to a large (de facto infinite) dataset, using physics simulation in the loop. In this way, we free humans from babysitting the bots all day. The future of robot data is generative. The future of the entire robot learning pipeline will also be generative. 🧵

I don’t know if we live in a Matrix, but I know for sure that robots will spend most of their lives in simulation. Let machines train machines. I’m excited to introduce DexMimicGen, a massive-scale synthetic data generator that enables a humanoid robot to learn complex skills from only a handful of human demonstrations. Yes, as few as 5! DexMimicGen addresses the biggest pain point in robotics: where do we get data? Unlike with LLMs, where vast amounts of texts are readily available, you cannot simply download motor control signals from the internet. So researchers teleoperate the robots to collect motion data via XR headsets. They have to repeat the same skill over and over and over again, because neural nets are data hungry. This is a very slow and uncomfortable process. At NVIDIA, we believe the majority of high-quality tokens for robot foundation models will come from simulation. What DexMimicGen does is to trade GPU compute time for human time. It takes one motion trajectory from human, and multiplies into 1000s of new trajectories. A robot brain trained on this augmented dataset will generalize far better in the real world. Think of DexMimicGen as a learning signal amplifier. It maps a small dataset to a large (de facto infinite) dataset, using physics simulation in the loop. In this way, we free humans from babysitting the bots all day. The future of robot data is generative. The future of the entire robot learning pipeline will also be generative. 🧵

Jim Fan

165,246 次观看 • 1 年前

Christine (郡主Christine (✱,✱)), co-founder of Axis Robotics spent eight years ago studying mechanical engineering and working with ABB and KUKA robots before pivoting through stablecoin payments and landing on physical AI data. Axis Robotics has raised $12M to build a distributed training engine where anyone with a browser can teach robots. The network already has 200,000+ contributors and a dataset of 50,000+ robot trajectories across 207 tasks. On today's episode: why noisy human data beats "perfect" AI data, why simulation might be the key to physical AI's GPT moment, and the uncomfortable truth about people getting paid to train the robots that'll eventually replace them. Timestamps: 0:00 Intro: Building the Data Layer Behind Physical AI 1:34 Hot Take: Noisy Data vs. Perfect Data 3:16 Why Simulation Data Will Bring the GPT Moment for Robotics 7:25 From Stablecoin Payments to Robotics Data 10:08 Why Axis Chose Data Over Building Robots 16:10 Live Demo: Training a Robot Arm Like Playing FIFA 24:02 How Contributors Get Paid (Tokens + Stablecoin) 26:42 Inside the $12M Seed Round Pitch 32:08 200K+ Contributors & Viral Growth via Base 38:38 Sim-to-Real: The Living Data Engine 41:34 Token & Web3 Plans for Axis Robotics 43:00 Rapid Fire Questions

Christine (郡主Christine (✱,✱)), co-founder of Axis Robotics spent eight years ago studying mechanical engineering and working with ABB and KUKA robots before pivoting through stablecoin payments and landing on physical AI data. Axis Robotics has raised $12M to build a distributed training engine where anyone with a browser can teach robots. The network already has 200,000+ contributors and a dataset of 50,000+ robot trajectories across 207 tasks. On today's episode: why noisy human data beats "perfect" AI data, why simulation might be the key to physical AI's GPT moment, and the uncomfortable truth about people getting paid to train the robots that'll eventually replace them. Timestamps: 0:00 Intro: Building the Data Layer Behind Physical AI 1:34 Hot Take: Noisy Data vs. Perfect Data 3:16 Why Simulation Data Will Bring the GPT Moment for Robotics 7:25 From Stablecoin Payments to Robotics Data 10:08 Why Axis Chose Data Over Building Robots 16:10 Live Demo: Training a Robot Arm Like Playing FIFA 24:02 How Contributors Get Paid (Tokens + Stablecoin) 26:42 Inside the $12M Seed Round Pitch 32:08 200K+ Contributors & Viral Growth via Base 38:38 Sim-to-Real: The Living Data Engine 41:34 Token & Web3 Plans for Axis Robotics 43:00 Rapid Fire Questions

Pukerainbow 🤮🌈

24,329 次观看 • 5 天前

NVIDIA just announced EgoScale 🤖🧠 NVIDIA Research has uncovered a log-linear scaling law for robot dexterity by pretraining VLA models on over 20,000 hours of egocentric human video This massive dataset is 20 times larger than previous efforts and proves that robot intelligence follows a predictable path: the more human data, the lower the loss The secret is a simple recipe combining large-scale human pretraining with a small amount of aligned human-robot mid-training to bridge the gap In testing, this method boosted the average success rate by 54% on a 22-DoF robotic hand compared to policies built without pretraining EgoScale also enables one-shot task adaptation and works across different hardware, suggesting that human motion is a universal motor prior for robots Website: Paper: Source: NVIDIA Research #Robot #Humanoid #Robotics #AI #EmbodiedAI #PhysicalAI #NVIDIA #EgoScale #GR00T

NVIDIA just announced EgoScale 🤖🧠 NVIDIA Research has uncovered a log-linear scaling law for robot dexterity by pretraining VLA models on over 20,000 hours of egocentric human video This massive dataset is 20 times larger than previous efforts and proves that robot intelligence follows a predictable path: the more human data, the lower the loss The secret is a simple recipe combining large-scale human pretraining with a small amount of aligned human-robot mid-training to bridge the gap In testing, this method boosted the average success rate by 54% on a 22-DoF robotic hand compared to policies built without pretraining EgoScale also enables one-shot task adaptation and works across different hardware, suggesting that human motion is a universal motor prior for robots Website: Paper: Source: NVIDIA Research #Robot #Humanoid #Robotics #AI #EmbodiedAI #PhysicalAI #NVIDIA #EgoScale #GR00T

RoboHub🤖

43,752 次观看 • 7 个月前

Wow! Super impressive work by the new Amazon FAR team (from Covariant acquisition). Mapping long sequences of human motion (>30 sec) on robots with a differing shapes or interating with objects (box, table, etc) of different size. Enabling easier in-simulation data-augmentation and zero-shoot transfer. Super impressive and huge help to reduce the need for human teleop data (which is very complex to gather for humanoids) Dataset trajectories on Hugging Face (search OmniRetarget), full code framework to come soon Project page has some pretty three.js interactive demos

Wow! Super impressive work by the new Amazon FAR team (from Covariant acquisition). Mapping long sequences of human motion (>30 sec) on robots with a differing shapes or interating with objects (box, table, etc) of different size. Enabling easier in-simulation data-augmentation and zero-shoot transfer. Super impressive and huge help to reduce the need for human teleop data (which is very complex to gather for humanoids) Dataset trajectories on Hugging Face (search OmniRetarget), full code framework to come soon Project page has some pretty three.js interactive demos

Thomas Wolf

149,831 次观看 • 1 年前

🔥 JUST IN: Open-source robotics dataset from 100% real-world scenarios! 🤯 Chinese robotics company AGIBOT just released AGIBOT WORLD 2026, an open-source dataset systematically covering key embodied AI research directions. Built entirely from real-world environments: commercial spaces, and homes. Collected using AGIBOT G2 robots in free-form collection mode, providing structured, accurately annotated, high-quality data. Digital twin technology creates 1:1 scale replicas in simulation matching the real environments. Both real-world and simulation data are open-sourced. The AGIBOT G2 platform collects multiple data types simultaneously: RGB(D) cameras, tactile sensors, force sensors, LiDAR, IMU, and full-body joint states. Whole-body control coordinates arms, waist, and hands for complex tasks. First-person teleoperation lets operators control the robot from its perspective. The tasks covered are fine-grained manipulation, ultra-long-horizon tasks, spatial navigation, dual-arm coordination, and multi-agent/human-robot collaboration. The dataset includes error-recovery trajectories with annotations. Most datasets only show successful demonstrations. AGIBOT includes failures and how the robot recovers, teaching models how to handle mistakes. After collection, data is tested through policy training and real-robot deployment to ensure quality. Then processed through industrial quality control with multiple screening and cleaning rounds. Making it open-source accelerates embodied AI research by giving researchers access to high-quality real-world robot data at scale. 🇨🇳 Learn more here: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

🔥 JUST IN: Open-source robotics dataset from 100% real-world scenarios! 🤯 Chinese robotics company AGIBOT just released AGIBOT WORLD 2026, an open-source dataset systematically covering key embodied AI research directions. Built entirely from real-world environments: commercial spaces, and homes. Collected using AGIBOT G2 robots in free-form collection mode, providing structured, accurately annotated, high-quality data. Digital twin technology creates 1:1 scale replicas in simulation matching the real environments. Both real-world and simulation data are open-sourced. The AGIBOT G2 platform collects multiple data types simultaneously: RGB(D) cameras, tactile sensors, force sensors, LiDAR, IMU, and full-body joint states. Whole-body control coordinates arms, waist, and hands for complex tasks. First-person teleoperation lets operators control the robot from its perspective. The tasks covered are fine-grained manipulation, ultra-long-horizon tasks, spatial navigation, dual-arm coordination, and multi-agent/human-robot collaboration. The dataset includes error-recovery trajectories with annotations. Most datasets only show successful demonstrations. AGIBOT includes failures and how the robot recovers, teaching models how to handle mistakes. After collection, data is tested through policy training and real-robot deployment to ensure quality. Then processed through industrial quality control with multiple screening and cleaning rounds. Making it open-source accelerates embodied AI research by giving researchers access to high-quality real-world robot data at scale. 🇨🇳 Learn more here: ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Lukas Ziegler

40,583 次观看 • 5 个月前

First there was Skild AI showing zero-shot transfer to unfamiliar homes, now Figure has a humanoid doing whole-body tasks across 30 homes it’s never seen before. Helix 2.5 was pretrained from scratch on Index, Figure’s human-behavior dataset, then fine-tuned for those three tasks elsewhere. Zero-shot meaning new homes and objects, not tasks it was never trained to do. No data collection or fine-tuning in those 30 homes. adding Index pretraining raised complete task success from 9% to 56%, with no partial credit. They also found that doubling the human pretraining data predictably improved robot-action prediction, enough to forecast their largest run’s loss to four decimal places. Obviously 56% isn’t perfect, but getting that much more generalization from human experience alone is pretty exciting.

First there was Skild AI showing zero-shot transfer to unfamiliar homes, now Figure has a humanoid doing whole-body tasks across 30 homes it’s never seen before. Helix 2.5 was pretrained from scratch on Index, Figure’s human-behavior dataset, then fine-tuned for those three tasks elsewhere. Zero-shot meaning new homes and objects, not tasks it was never trained to do. No data collection or fine-tuning in those 30 homes. adding Index pretraining raised complete task success from 9% to 56%, with no partial credit. They also found that doubling the human pretraining data predictably improved robot-action prediction, enough to forecast their largest run’s loss to four decimal places. Obviously 56% isn’t perfect, but getting that much more generalization from human experience alone is pretty exciting.

Chris

15,972 次观看 • 14 天前

LP-MusicCaps: LLM-Based Pseudo Music Captioning paper page: Automatic music captioning, which generates natural language descriptions for given music tracks, holds significant potential for enhancing the understanding and organization of large volumes of musical data. Despite its importance, researchers face challenges due to the costly and time-consuming collection process of existing music-language datasets, which are limited in size. To address this data scarcity issue, we propose the use of large language models (LLMs) to artificially generate the description sentences from large-scale tag datasets. This results in approximately 2.2M captions paired with 0.5M audio clips. We term it Large Language Model based Pseudo music caption dataset, shortly, LP-MusicCaps. We conduct a systemic evaluation of the large-scale music captioning dataset with various quantitative evaluation metrics used in the field of natural language processing as well as human evaluation. In addition, we trained a transformer-based music captioning model with the dataset and evaluated it under zero-shot and transfer-learning settings. The results demonstrate that our proposed approach outperforms the supervised baseline model.

LP-MusicCaps: LLM-Based Pseudo Music Captioning paper page: Automatic music captioning, which generates natural language descriptions for given music tracks, holds significant potential for enhancing the understanding and organization of large volumes of musical data. Despite its importance, researchers face challenges due to the costly and time-consuming collection process of existing music-language datasets, which are limited in size. To address this data scarcity issue, we propose the use of large language models (LLMs) to artificially generate the description sentences from large-scale tag datasets. This results in approximately 2.2M captions paired with 0.5M audio clips. We term it Large Language Model based Pseudo music caption dataset, shortly, LP-MusicCaps. We conduct a systemic evaluation of the large-scale music captioning dataset with various quantitative evaluation metrics used in the field of natural language processing as well as human evaluation. In addition, we trained a transformer-based music captioning model with the dataset and evaluated it under zero-shot and transfer-learning settings. The results demonstrate that our proposed approach outperforms the supervised baseline model.

AK

78,794 次观看 • 3 年前

A big part of scaling robot learning to solve real-world problems is that we somehow need to get enough diverse, high-quality data to train our robots to perform useful things. GPT and its fellow large language models were bootstrapped and proved out on a massive dataset of real-world language data. Unfortunately, despite our best efforts, similarly massive datasets don’t really exist for robotics — so, in our unending pursuit of high-quality, useful data, we turn to simulation. I compared a couple recent works on sim-to-real robot manipulation, which discuss how to train perception-driven manipulation policies in simulation, in such a way that they’re useful in the real world. - DextraH-RGB, from NVIDIA - Sim-and-Real Co-Training: A Simple Recipe for Vision-Based Robotic Manipulation, also from NVIDIA — specifically the GEAR lab - Sim-to-Real Reinforcement Learning for Vision-Based Dexterous Manipulation on Humanoids, another GEAR lab paper - Local Policies Enable Zero-shot Long-Horizon Manipulation, from CMU (video from DextrAH-RGB)

A big part of scaling robot learning to solve real-world problems is that we somehow need to get enough diverse, high-quality data to train our robots to perform useful things. GPT and its fellow large language models were bootstrapped and proved out on a massive dataset of real-world language data. Unfortunately, despite our best efforts, similarly massive datasets don’t really exist for robotics — so, in our unending pursuit of high-quality, useful data, we turn to simulation. I compared a couple recent works on sim-to-real robot manipulation, which discuss how to train perception-driven manipulation policies in simulation, in such a way that they’re useful in the real world. - DextraH-RGB, from NVIDIA - Sim-and-Real Co-Training: A Simple Recipe for Vision-Based Robotic Manipulation, also from NVIDIA — specifically the GEAR lab - Sim-to-Real Reinforcement Learning for Vision-Based Dexterous Manipulation on Humanoids, another GEAR lab paper - Local Policies Enable Zero-shot Long-Horizon Manipulation, from CMU (video from DextrAH-RGB)

Chris Paxton

20,486 次观看 • 1 年前

Synthetic data will provide the next trillion tokens to fuel our hungry models. I'm excited to announce MimicGen: massively scaling up data pipeline for robot learning! We multiply high-quality human data in simulation with digital twins. Using 50,000 training episodes across 18 tasks, multiple simulators, and even in the real-world! The idea is simple: 1. Humans tele-operate the robot to complete a task. It is extremely high-quality but also very slow and expensive. 2. We create a digital twin of the robot and the scene in high-fidelity, GPU-accelerated simulation. 3. We can now move objects around, replace with new assets, and even change the robot hand - basically augment the training data with procedural generation. 4. Export the successful episodes, and feed that to a neural network! You now have an near-infinite stream of data. One of the key reasons that robotics lags far behind other AI fields is the lack of data: you cannot scrape control signals from the internet. They simply don't exist in-the-wild. MimicGen shows the power of synthetic data and simulation to keep our scaling laws alive. I believe this principle apply beyond robotics. We are quickly exhausting the high-quality, real tokens from the web. Artificial intelligence from artificial data will be the way forward. We are big fans of the OSS community. As usual, we open-source everything, including the generated dataset! - Website: - Paper: - Dataset is hosted on HuggingFace (thanks AK!!): - Code: MimicGen is led by Ajay Mandlekar, deep dive in the thread:

Synthetic data will provide the next trillion tokens to fuel our hungry models. I'm excited to announce MimicGen: massively scaling up data pipeline for robot learning! We multiply high-quality human data in simulation with digital twins. Using 50,000 training episodes across 18 tasks, multiple simulators, and even in the real-world! The idea is simple: 1. Humans tele-operate the robot to complete a task. It is extremely high-quality but also very slow and expensive. 2. We create a digital twin of the robot and the scene in high-fidelity, GPU-accelerated simulation. 3. We can now move objects around, replace with new assets, and even change the robot hand - basically augment the training data with procedural generation. 4. Export the successful episodes, and feed that to a neural network! You now have an near-infinite stream of data. One of the key reasons that robotics lags far behind other AI fields is the lack of data: you cannot scrape control signals from the internet. They simply don't exist in-the-wild. MimicGen shows the power of synthetic data and simulation to keep our scaling laws alive. I believe this principle apply beyond robotics. We are quickly exhausting the high-quality, real tokens from the web. Artificial intelligence from artificial data will be the way forward. We are big fans of the OSS community. As usual, we open-source everything, including the generated dataset! - Website: - Paper: - Dataset is hosted on HuggingFace (thanks AK!!): - Code: MimicGen is led by Ajay Mandlekar, deep dive in the thread:

Jim Fan

332,238 次观看 • 2 年前

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

467,416 次观看 • 1 年前

Quadrupeds are fast. Agile. Great at locomotion. But can they manipulate? A new approach from Carnegie Mellon University, Google DeepMind, and Bosch is teaching quadrupedal robots to do more than walk, they’re learning to interact. It’s called Human2LocoMan: a system that uses human data to pretrain robot policies before finetuning on real hardware. The result? A four-legged robot that can walk, carry, organize, scoop, and sort with both single and dual-arm control. By pretraining on human motion, they cut the amount of robot data in half—while improving success rates by over 80% in unfamiliar environments. Their Modularized Cross-Embodiment Transformer (MXT) learns from both human and robot demonstrations, then generalizes those skills to physical tasks—no hardcoded behaviors required. It’s locomotion and manipulation. A quadruped that can walk and clean up after itself?

Quadrupeds are fast. Agile. Great at locomotion. But can they manipulate? A new approach from Carnegie Mellon University, Google DeepMind, and Bosch is teaching quadrupedal robots to do more than walk, they’re learning to interact. It’s called Human2LocoMan: a system that uses human data to pretrain robot policies before finetuning on real hardware. The result? A four-legged robot that can walk, carry, organize, scoop, and sort with both single and dual-arm control. By pretraining on human motion, they cut the amount of robot data in half—while improving success rates by over 80% in unfamiliar environments. Their Modularized Cross-Embodiment Transformer (MXT) learns from both human and robot demonstrations, then generalizes those skills to physical tasks—no hardcoded behaviors required. It’s locomotion and manipulation. A quadruped that can walk and clean up after itself?

Lukas Ziegler

61,610 次观看 • 1 年前

Figure announces Project Go-Big: Internet-Scale Humanoid Pretraining and Direct Human-to-Robot Transfer The company has achieved a new milestone for the Helix AI model: after training exclusively on egocentric human data, Figure robots can now navigate real-world spaces from natural language commands such as “Go water the plants.” This also enabled a unified Helix model that can output both manipulation and navigation commands to the continuous control system. This ties into the Brookfield partnership that they announced yesterday. Brookfield is a global asset manager overseeing $1 trillion in assets, including 100,000 residential units, 500M square feet of commercial office space, and 160M square feet of logistics space. Using human video capture in a variety of Brookfield environments, Figure will amass critical pretraining data at scale, allowing them to explore the full potential of internet-scale embodied datasets for locomotion and manipulation.

Figure announces Project Go-Big: Internet-Scale Humanoid Pretraining and Direct Human-to-Robot Transfer The company has achieved a new milestone for the Helix AI model: after training exclusively on egocentric human data, Figure robots can now navigate real-world spaces from natural language commands such as “Go water the plants.” This also enabled a unified Helix model that can output both manipulation and navigation commands to the continuous control system. This ties into the Brookfield partnership that they announced yesterday. Brookfield is a global asset manager overseeing $1 trillion in assets, including 100,000 residential units, 500M square feet of commercial office space, and 160M square feet of logistics space. Using human video capture in a variety of Brookfield environments, Figure will amass critical pretraining data at scale, allowing them to explore the full potential of internet-scale embodied datasets for locomotion and manipulation.

The Humanoid Hub

32,953 次观看 • 1 年前

$*New Paper on AI & Democracy* Imagine two approaches to democracy. The one we have today, where citizens choose a professional politician to represent them and others. Or an augmented form of democracy, where each citizen controls a personalized AI that helps them participate in thousands of nuanced decisions. This second approach is the idea of Augmented Democracy I introduced six years ago at TED. In our latest paper we explore a simplified version of Augmented Democracy by combining off-the-shelf LLMs, such as ChatGPT, with data collected using a collaborative government program builder. This was an online game where people build a personalized government program using proposals extracted from the programs of the candidates of the 2022 presidential election in Brazil. So how accurate are these augmented forms of democracy? Imagine a user who gave us 40 answers. We can use the first 20 to fine-tune a model that we can test using the 20 answers the model didn’t see. We can then compare the accuracy of these predictions with the ones obtained by a “bundle” rule, which assumes that users that self-reported to be from the left or right always chose the proposals from the candidate that shares their political identity. This showed us that LLMs were more accurate at predicting policy preferences than the bundle rule, meaning that the preferences captured in the participation data were more nuanced than a left-right axis, and that the LLMs can capture some of that nuance. Also, the LLMs can choose among policies coming from the same candidate, which is something that we cannot do using a bundle rule. But can these LLMs help us complete the aggregate preferences of the population? Direct or unbundled forms of participation can result in incomplete data when people answer only a fraction of all questions. In our paper, we simulate this incompleteness by sampling the full dataset. We ask how close we can get to the full dataset by using a random sample, or a random sample augmented by predictions made by these LLMs. Overall, we find that LLM-augmented data gets much closer to the full dataset than a pure random sample. These results do not mean that augmented democracy technology is ready, but they means we are in a much better place to continue exploring this idea than six years ago. This paper was a collaborative effort with Jairo Gudino, PhD student at CCL at the University of Toulouse Capitole and Umberto Grandi from IRIT also at the University of Toulouse Capitole. We hope you find these results insightful!$

New Paper on AI & Democracy Imagine two approaches to democracy. The one we have today, where citizens choose a professional politician to represent them and others. Or an augmented form of democracy, where each citizen controls a personalized AI that helps them participate in thousands of nuanced decisions. This second approach is the idea of Augmented Democracy I introduced six years ago at TED. In our latest paper we explore a simplified version of Augmented Democracy by combining off-the-shelf LLMs, such as ChatGPT, with data collected using a collaborative government program builder. This was an online game where people build a personalized government program using proposals extracted from the programs of the candidates of the 2022 presidential election in Brazil. So how accurate are these augmented forms of democracy? Imagine a user who gave us 40 answers. We can use the first 20 to fine-tune a model that we can test using the 20 answers the model didn’t see. We can then compare the accuracy of these predictions with the ones obtained by a “bundle” rule, which assumes that users that self-reported to be from the left or right always chose the proposals from the candidate that shares their political identity. This showed us that LLMs were more accurate at predicting policy preferences than the bundle rule, meaning that the preferences captured in the participation data were more nuanced than a left-right axis, and that the LLMs can capture some of that nuance. Also, the LLMs can choose among policies coming from the same candidate, which is something that we cannot do using a bundle rule. But can these LLMs help us complete the aggregate preferences of the population? Direct or unbundled forms of participation can result in incomplete data when people answer only a fraction of all questions. In our paper, we simulate this incompleteness by sampling the full dataset. We ask how close we can get to the full dataset by using a random sample, or a random sample augmented by predictions made by these LLMs. Overall, we find that LLM-augmented data gets much closer to the full dataset than a pure random sample. These results do not mean that augmented democracy technology is ready, but they means we are in a much better place to continue exploring this idea than six years ago. This paper was a collaborative effort with Jairo Gudino, PhD student at CCL at the University of Toulouse Capitole and Umberto Grandi from IRIT also at the University of Toulouse Capitole. We hope you find these results insightful!

César A. Hidalgo

26,915 次观看 • 1 年前

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data. 2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the motion of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro -> RoboCasa produces N (varying visuals) -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are building tools to enable everyone in the ecosystem to scale up with us. Links in thread:

Jim Fan

364,670 次观看 • 2 年前

$QVAC SDK will support in 0.9.0 (gonna be release in ~10 days) LoRA fine-tuning directly on-device, letting developers customize LLMs with their own data without sending anything to the cloud. You just load a base model, point it at your training dataset, and get a lightweight LoRA adapter back — all running locally. The fine-tuned model can then be used for inference immediately, with no extra setup. Why it matters: LoRA (Low-Rank Adaptation) fine-tuning lets you specialize a general-purpose language model for your specific use case — like matching a brand's tone, mastering domain terminology, or following a particular output format — using a fraction of the compute a full fine-tune would require. QVAC handles the entire workflow locally: dataset preparation, training with configurable hyperparameters, checkpoint saving, and seamless inference with the resulting adapter. Your data never leaves the device. The developer experience: Fine-tuning with QVAC is as simple as calling "sdk.finetune()" with your dataset and a few hyperparameters. Training runs entirely on your local hardware, produces a compact LoRA adapter file, and supports pause/resume so you can stop a job and pick it back up without losing progress. The result plugs straight into QVAC's inference pipeline — no model conversion, no deployment step, just immediate local completions with your fine-tuned model.$

QVAC SDK will support in 0.9.0 (gonna be release in ~10 days) LoRA fine-tuning directly on-device, letting developers customize LLMs with their own data without sending anything to the cloud. You just load a base model, point it at your training dataset, and get a lightweight LoRA adapter back — all running locally. The fine-tuned model can then be used for inference immediately, with no extra setup. Why it matters: LoRA (Low-Rank Adaptation) fine-tuning lets you specialize a general-purpose language model for your specific use case — like matching a brand's tone, mastering domain terminology, or following a particular output format — using a fraction of the compute a full fine-tune would require. QVAC handles the entire workflow locally: dataset preparation, training with configurable hyperparameters, checkpoint saving, and seamless inference with the resulting adapter. Your data never leaves the device. The developer experience: Fine-tuning with QVAC is as simple as calling "sdk.finetune()" with your dataset and a few hyperparameters. Training runs entirely on your local hardware, produces a compact LoRA adapter file, and supports pause/resume so you can stop a job and pick it back up without losing progress. The result plugs straight into QVAC's inference pipeline — no model conversion, no deployment step, just immediate local completions with your fine-tuned model.

Paolo Ardoino 🤖

42,413 次观看 • 5 个月前