Chonghao Sima @chonghaosima.bsky.social · 11/02/2026🏆 Looking ahead: We are working on hosting a competition in 2026. We want to see different policies and hardware setups compete head-to-head in the same arena. Let's put them to the real test! 000
Chonghao Sima @chonghaosima.bsky.social · 11/02/2026We’ve also released some "Quality of Life" tools to streamline your workflow: ⚡ Ultra-fast compute norm state (Significantly faster than the official LeRobot implementation!) 🛠️ Micro-tools for LeRobot dataset manipulation 🎮 DAgger support 000
Chonghao Sima @chonghaosima.bsky.social · 11/02/2026Our repo includes code, data, hardware manuals, and inference setups for AgileX (Songling) and Ark (rolling out gradually). We really hope to bring the reproducibility standards of the CV community into this space. 🔧 000
Chonghao Sima @chonghaosima.bsky.social · 11/02/2026Huge shoutout to the entire team and everyone involved behind the scenes! 🙌 KAI0 is going fully open-source this week. 🚀 📄 Paper: arxiv.org/abs/2602.09021 💻 Code: github.com/OpenDriveLab...arxiv.org$χ_{0}$: Resource-Aware Robust Manipulation via Taming Distributional InconsistenciesHigh-reliability long-horizon robotic manipulation has traditionally relied on large-scale data and compute to understand complex real-world dynamics. However, we identify that the primary bottleneck ... 300
Chonghao Sima @chonghaosima.bsky.social · 24/12/2025[5/5] Bottom Line • Not all robot data is equally valuable • Fast iteration > bruteforce scaling • Weight-space merging can outperform joint training • Stage-aware advantage estimation helps long-horizon tasks 📄 Full report: Q1 2026 📦 Data + checkpoints + challenge: 2026 000
Chonghao Sima @chonghaosima.bsky.social · 24/12/2025[4/5] Problem: Long-Horizon Credit Assignment 6-minute tasks. Which actions actually helped? Solution → Stage Advantage: • Decompose into semantic stages • Predict advantage directly (not value-diff) • Smoother supervision, less error compounding 100
Chonghao Sima @chonghaosima.bsky.social · 24/12/2025[3/5] Problem: Expensive Iteration Collect new data → Retrain everything → Repeat Slow yet expensive. How? Model Arithmetic: • Train only on new data • Merge via weight interpolation • Merged model > full-dataset model Models trained separately preserve distinct modes. 100
Chonghao Sima @chonghaosima.bsky.social · 24/12/2025[2/5] Problem: Distribution Mismatch Training data ≠ Model behavior ≠ Real-world execution This gap causes failures. Solution → Mode Consistency: • DAgger for failure recovery • Augmentation for coverage • Inference smoothing for clean execution 100
Chonghao Sima @chonghaosima.bsky.social · 24/12/2025🧥 Live-stream robotic teamwork that folds clothes. 6 clothes in 3 minutes straight. χ₀ = 20hrs data + 8 A100s + 3 key insights: - Mode Consistency: align your distributions - Model Arithmetic: merge, don't retrain - Stage Advantage: pivot wisely 🔗 mmlab.hk/research/kai0 checkout 3mins demo 161
Chonghao Sima @chonghaosima.bsky.social · 10/06/2025🚀 HERE WE GO! Join us at CVPR 2025 for a full-day tutorial: “Robotics 101: An Odyssey from a Vision Perspective” 🗓️ June 12 • 📍 Room 202B, Nashville Meet our incredible lineup of speakers covering topics from agile robotics to safe physical AI at: opendrivelab.com/cvpr2025/tut... #cvpr2025 162
Chonghao Sima @chonghaosima.bsky.social · 08/06/2025Thanks for sharing! I will host the workshop for the whole day and welcome anyone who is struggling with current embodied AI trend to visit and chat and exchange ideas! We want to hear the opposite opinions from vision and robotics people on the topic of autonomy. 030
Reposted by Chonghao SimaAndrei Bursuc @abursuc.bsky.social · 07/06/2025When at @cvprconference.bsky.social a major challenge is how to split yourself for super amazing workshops. I'm afraid to announce that w/ our workshop on "Embodied Intelligence for Autonomous Systems on the Horizon" we will make this choice even harder: opendrivelab.com/cvpr2025/wor... #cvpr2025 1144
Chonghao Sima @chonghaosima.bsky.social · 13/04/2025Wonderful end-to-end driving benchmark! We are getting **closer and closer** to **close-loop** evaluation in real world! 010
Chonghao Sima @chonghaosima.bsky.social · 24/03/2025@katrinrenz.bsky.social @kashyap7x.bsky.social @andreasgeiger.bsky.social @hongyang.bsky.social @opendrivelab.bsky.social 021
Chonghao Sima @chonghaosima.bsky.social · 24/03/2025DriveLM got 1k stars on GitHub, my first project reaching such milestone. Great thanks to all my collaborators who contribute much to this project, many thanks to the community who participate and contribute better insight upon this dataset, and wish this is not my end! 1102
Chonghao Sima @chonghaosima.bsky.social · 17/03/2025Fun fact: the second character in my last name is 🐎 as well. 030
Chonghao Sima @chonghaosima.bsky.social · 17/03/2025Thanks for sharing! We long to know if we could improve e2e planner with limited but online data and compute, as performance with more training data seems plateau. However, online failure cases are unexplored as they couldn’t directly contribute to the model performance via previous training scheme. 110
Chonghao Sima @chonghaosima.bsky.social · 06/03/2025Random thoughts today: in humanoid research the methodology is basically decided by the final tasks/demo you would like to show off. 010
Reposted by Chonghao SimaOpenDriveLab @opendrivelab.bsky.social · 05/03/2025🌟 Previewing the UniAD 2.0 🚀 A milestone upgrade on the codebase of the #CVPR2023 best paper UniAD. 👉 Check out this branch github.com/OpenDriveLab..., and we will get you more details soon 093
Reposted by Chonghao SimaOpenDriveLab @opendrivelab.bsky.social · 03/03/2025🚀 This year, we’re bringing you three thrilling tracks in Embodied AI and Autonomous Driving, with a total prize pool of $100,000! Now get ready and join the competition! Visit the challenge website: opendrivelab.com/challenge2025 And more on #CVPR2025: opendrivelab.com/cvpr2025 054
Chonghao Sima @chonghaosima.bsky.social · 30/12/2024Thanks for all the staff who work hard to make it happen! Love to hear your feedback. 010
Chonghao Sima @chonghaosima.bsky.social · 11/12/2024For 1, We may need a "greatest common divisor" among tasks/algorithms/embodiments. For 2, retargetting seems to be the most critical issue. For 3, should we follow sample-efficiency RL or VLM-based e2e methods? 000
Chonghao Sima @chonghaosima.bsky.social · 11/12/2024Random thoughts (again) on: 1. Benchmark & Evaluation & Metrics 2. Data collection (especially tele-op) 3. Policy network architecture & training receipt. 100
Chonghao Sima @chonghaosima.bsky.social · 04/12/2024Random thoughts today: situation in humanoids today is similar to autonomous driving back into 2020-ish. Different hardware setups, people more favor of RL-based planning and sim2real deployment, etc. Will humanoids get into a similar development curve like driving? 030
Reposted by Chonghao Simaemilk @ernerfeldt.bsky.social · 02/12/2024We implemented undo in @rerun.io by storing the viewer state in the same type of in-memory database we use for the recorded data. Have a look (sound on!) 34710