正在加载视频...

视频加载失败

Autoregressive diffusion models drift for long videos? 📉 We fixed it.🚀 Speed + Stability = ✅ Meeting *Test-Time Correction (TTC)*. We stop error accumulation in its tracks without any retraining. ✅ Training-free ✅ 1 minute+ stable generation ✅ Negligible overhead

16,460 次观看 • 4 个月前 •via X (Twitter)

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

A team tested Pi0, Pi0 Fast, Gr00t, and ACT on real robot arms in manufacturing tasks. (🔖 Bookmark this for later!) The task was precise: place thin rectangular frames from a messy stack into a holder. The team fine-tuned each model on 100 real trajectories and compared training time, inference speed, motion quality, and success rates. ⬇️ Here’s a breakdown of what they found Pi0 (Original) ✅ Strongest overall performance in precise pick-and-place ✅ High success rate even in edge cases ✅ Longest training time (~11 hours, ~$30 per run) ✅ Inference time of 80 ms causes short pauses between actions Despite delays, it handles complex scenarios well… solid for high-precision tasks, but slow to train. Gr00t ✅ Trains fast (~2 hours, ~$5 per run) ✅ Performs almost as well as Pi0 on large-object tasks ✅ Struggles with fine precision; random movement in some trials ✅ More training didn’t fix jitter or random offsets Best suited for tasks where exact precision isn’t critical. Not ready for manufacturing-grade accuracy without more tuning. Pi0 Fast ✅ Promised faster training, but results were underwhelming ✅ Training at 6 hours still showed low success rates ✅ Inference was slower than expected ✅ Not reliable for generalizing even slightly new tasks Currently too unstable for real-world deployment. Doesn’t live up to the “Fast” name yet. ACT (Baseline) ✅ 200MB model—lightweight, but limited ✅ Struggles with stacked objects or ambiguous scenes ✅ Success rates around 70% in best-case setups ✅ Can’t match newer models on precision or generalization Still a solid baseline, but clearly a generation behind in robustness. 🚨 Extra Notes All newer models share a common issue: •Inference takes longer than a frame (80 ms vs 33 ms), so robots “pause” between chunks. •This results in jittery movements, but not a dealbreaker unless tasks are time-sensitive. Language-conditioned tasks also fell short: after training on two labeled tasks, the model couldn’t generalize to a third unseen combination using only text prompts. ✅ The good news? These models adapt well to new robot arms with quick fine-tuning. ❌ The bad news? There’s still no plug-and-play solution for improving performance after deployment. Reinforcement learning or DAgger-style data collection during real-world operation may be the next big step, something many teams in robotics are actively working on.

Ilir Aliu

21,844 次观看 • 1 年前

I’m thrilled to announce that we just released GraspGen, a multi-year project we have been cooking at NVIDIA Robotics 🚀 GraspGen: A Diffusion-Based Framework for 6-DOF Grasping Grasping is a foundational challenge in robotics 🤖 — whether for industrial picking or general-purpose humanoids. VLA + real data collection is all the rage now but is expensive and scales poorly for this task. For every new gripper and/or scene, you’ll have to recollect the dataset in this paradigm for the best perf. 💡Key Idea: Since grasping is such a well-defined task in simulation - why can’t we just scale synthetic data generation and train a generative model for grasping? By embracing modularity and standardized grasp formats, we can make this a turnkey technology that works zero-shot for multiple settings. GraspGen is a modular framework for diffusion-based 6-DOF grasp generation that scales across embodiment types, observability conditions, clutter, task complexity. Key Features: ✅ Multi-embodiment support: suction, parallel-jaw, and multi-fingered grippers ✅ Generalization to partial + complete 3D point clouds ✅ Generalization to single-objects + cluttered scenes ✅ Modular design uses other robotics modules and foundation models (SAM2, cuRobo, FoundationStereo, FoundationPose). This allows GraspGen to focus on only one thing - grasp generation ✅ Training recipe: grasp discriminator is trained with On-Generator data from the diffusion model - so that it learns to correct the mistakes (if any) of the diffusion generator ✅ Real-time performance (~20 Hz) before any GPU acceleration; low memory footprint 📊 Results: • SOTA on the FetchBench [Han et al. CoRL 2024] benchmark • Zero-shot sim-to-real transfer on unknown objects and cluttered scenes • Dataset of 53M simulated grasps across 8K objects from Objaverse 📄 arXiv: 🌐 Website: 💻 Code: A huge thank you to everyone involved in this journey — excited to see what the community builds on top of it! Joint work with Clemens Eppner , Balakumar Sundaralingam , Yu-Wei, Jun Yamada Wentao Yuan and other collaborators #robotics #diffusionmodels #physicalAI #simtoreal

Adithya Murali

23,841 次观看 • 11 个月前

🚨 Keith Neumeyer: This Silver Rally is DIFFERENT WHY THIS SILVER RALLY IS DIFFERENT ✅ 2011 move: "Paper-driven short covering" ✅ 2024 move: "ALL PHYSICAL demand driven" ✅ "People waking up to silver as critical mineral" ✅ "This metal is required for everything - we can't travel, drive, or operate homes without it" THE STRUCTURAL DEFICIT REALITY ✅ 5 consecutive years of silver deficits ✅ Total deficit: ~1 BILLION ounces over 5 years ✅ Mine production: 850M oz/year | Consumption: 1.2B oz/year ✅ "These ounces are coming from investment hoards - that will end" WHY MINERS CAN'T SAVE US ✅ "Takes 3 years to drive tunnels to new discoveries" ✅ Mill upgrades require "years of work" ✅ No major silver mines coming online ✅ "We're not going to solve this at $50 silver" NEW DEMAND DRIVERS EMERGING ✅ India: 75M ounces imported recently ✅ AI data centers: "How are you going to build them without silver?" ✅ Nuclear renaissance: 30+ plants planned - all require silver PRICE PREDICTION & OUTLOOK ✅ "We're destined to go through new highs" ✅ "Wouldn't be surprised at $60-65 by year-end" ✅ Previous $40 prediction already shattered ✅ "This correction is healthy - settling before next leg up" According to one of silver's most respected CEOs, we're in a fundamentally different bull market driven by physical consumption that miners simply cannot meet - and the structural deficit means higher prices are inevitable, not speculative. HT: Kai Hoffmann Keith Neumeyer First Majestic #Silver #KeithNeumeyer #FirstMajestic #SilverSqueeze #PhysicalSilver #SupplyDeficit #Mining #CriticalMinerals #Investing

Mark

82,890 次观看 • 7 个月前

Is $SKR ready for its next leg up… or do we cool off first? Over 50K wallets have claimed SKR. Price is (currently) consolidating around $0.035 - $0.04 Here’s the 2 most likely scenarios 🧵👇 1️⃣ Current Setup: Holding Strong ✅ $SKR has unlocked for tens of thousands of wallets 👉 Yet we haven’t seen any major dump ✅ Consolidation around $0.04 👉 That’s 4x above launch levels 👉 Despite a massive unlock wave ✅ This shows real strength 👉 Price discovery is ongoing 👉 But downside pressure is limited so far 2️⃣ Scenario A: Push Higher ✅ Fresh upside needs real catalysts 👉 Launch hype usually only lasts for a few days ✅ Look out for bullish updates like: 👉 dApp integrations or new utility 👉 Seeker ecosystem incentives ✅ Bullish signal: 👉 Huge % of wallets already staked 👉 Reduces liquid supply & sell pressure 👉 Long-term holders ≠ paper hands 3️⃣ Scenario B: Pullback First ✅ SKR ran from $0.009 → $0.04 in <48h 👉 A short-term reset would be logical ✅ Without new announcements: 👉 Momentum may cool off 👉 People might take profit ✅ Imo, retrace into $0.02–$0.03 wouldn’t be bearish 👉 It’s just a healthier base before next move 4️⃣ Bullish Long-Term Either Way ✅ This wasn’t a random meme launch 👉 Solana Mobile planned the SKR rollout in phases ✅ Key unlocks, utility, & campaigns are still to come ✅ Even if price cools short-term: 👉 More news is almost certainly queued 👉 Seeker ecosystem has just started activating What do you think? Are we consolidating at $0.035 - $0.04 to push higher? Or do we first build a base at a lower range before new highs?

Marino

16,109 次观看 • 5 个月前