(1/N) How close are we to enabling robots to solve the long-horizon, complex tasks that matter in everyday life? 🚨 We are thrilled to invite you to join the 1st BEHAVIOR Challenge @NeurIPS 2025, submission deadline: 11/15. 🏆 Prizes: 🥇 $1,000 🥈 $500 🥉 $300

(2/N) In this challenge, take on 50 long-horizon mobile manipulation tasks that require diverse and complex low-level skills, powered by 1,200 hours of high-quality demonstrations. 🔗 Challenge website:

(3/N) As a reminder, BEHAVIOR is an open-source benchmark built on top of NVIDIA’s Omniverse, designed to enable and evaluate embodied AI and robotics solutions. It includes 1,000 everyday household tasks grounded in human needs. 📄 Paper: What’s new in this challenge?

(4/N) 🔍Feature #1: Large-Scale Demonstration Dataset • 50 tasks, 10,000 demos, a total of ~1,200 hours of data • Subtask and skill (30+) segmentation • Spatial relation annotation • Multi-granularity language annotation

(5/N) 💯Feature #2: High-Quality Data ✔️Teleoperated using our JoyLo interface ✔️Near-optimal, clean demos ✔️Consistent manipulation behaviors ✔️Moderate, consistent teleoperation speed 🚫 No sudden accelerations/decelerations 🚫 No failed grasps 🚫 No unintended collisions

(6/N) 🏠Feature #3: Long-Horizon Mobile Manipulation in Realistic Homes • Task durations range from 1 to 25 minutes (average 6.6 minutes) • Performed in household-scale scenes • Requires memory, planning, and reasoning over a long period of time

(7/N) 🧩Feature #4: Diverse State Transitions and Manipulation Skills • Spatial: next_to, inside, on_top, under, touching • Particles: covered, uncovered • Thermal: hot, cooked, on_fire, frozen • Others: open, closed, on, off, attached, sliced, diced

(8/N) 📊Provided Baselines We include a set of baselines to kickstart your experiments: • Classic behavioral cloning models: Diffusion Policy, WB-VIMA, ACT, BC-RNN • Pre-trained VLA models: OpenVLA, π_0 @physical_int

(9/N) 🧪Evaluation & Submission Submission instructions and evaluation details are available on our website: Start experimenting today and get ready to compete! Deadline: Nov. 15th Winners announced: Dec. 1st NeurIPS challenge: Dec. 6-7, San Diego, CA

(10/N) Together, let’s explore: ❓How close are we to solving long-horizon, complex, human-centric tasks? 🔀How to efficiently combine low-level control and high-level planning? 📉What are the generalization limits of current models? 📈 Are there scaling laws for embodied AI?

(11/N) 💬Join our Discord to ask questions and discuss: 🔗 We will also hold office hours on Monday and Thursday, 4:30-6pm PST over Zoom (see our website for the link). Whether you’re a robotics veteran or just entering the field, we’re here to support you.

(12/N) Proud of the amazing work from my students and collaborators: @Hang_Yin_ @wensi_ai @josiah_is_wong @cgokmenAI @ChengshuEricLi @YunfanJiang @mengdixu_ @EvansXuHan @sanjana__z @RavenHuang4 @RuohanZhang76 @jiajunwu_cs …

@Hang_Yin_ @wensi_ai @josiah_is_wong @cgokmenAI @ChengshuEricLi @YunfanJiang @mengdixu_ @EvansXuHan @sanjana__z @RavenHuang4 @RuohanZhang76 @jiajunwu_cs (13/N) And with the strong support from @wenlong_huang @chenwang_j @ArpitBahety @jiang_hanxiao @alexzhang_robo Niklas Vainio @RobobertoMM @YunzhuLiYZ @ManlingLi_ @Weiyu_Liu_ @silviocinguetta Karen Liu @hyogweon

@Hang_Yin_ @wensi_ai @josiah_is_wong @cgokmenAI @ChengshuEricLi @YunfanJiang @mengdixu_ @EvansXuHan @sanjana__z @RavenHuang4 @RuohanZhang76 @jiajunwu_cs @wenlong_huang @chenwang_j @ArpitBahety @jiang_hanxiao @alexzhang_robo @RobobertoMM @YunzhuLiYZ @ManlingLi_ @Weiyu_Liu_ @silviocinguetta @hyogweon (14/N) We thank @SimovationInc for providing high-quality JoyLo teleoperation data in simulation, reflecting their deep expertise in simulation and data quality. BEHAVIOR is built upon @nvidia’s Omniverse. We thank @nvidia for their continuous support.

@Hang_Yin_ @wensi_ai @josiah_is_wong @cgokmenAI @ChengshuEricLi @YunfanJiang @mengdixu_ @EvansXuHan @sanjana__z @RavenHuang4 @RuohanZhang76 @jiajunwu_cs @wenlong_huang @chenwang_j @ArpitBahety @jiang_hanxiao @alexzhang_robo @RobobertoMM @YunzhuLiYZ @ManlingLi_ @Weiyu_Liu_ @silviocinguetta @hyogweon @SimovationInc @nvidia (N/N) We thank our sponsors for their generous support: @SimovationInc @IMDAsg @StanfordHAI @SchmidtFutures

Hi @drfeifei, these prize values seem too low. Consider it doubled on each category from me. Happy to fund the progress of long horizon robotics.

thank you for the kind offer. the true prize is priceless - the quest for spatial and embodied intelligence. 😍

👍🏾🚀

i thought not including failed grasps and recovering from them was a failure point of current robotic datasets (as the robot won't learn how to recover from mistakes it'll make, but we wont) is that a wrong assumption?

If you're a researcher looking to collaborate, please don't hesitate to reach out.

prizes are.... what? lol

wow

Amazing work by Prof. Fei-Fei Li and team! 🚀 We’re so excited to see the 1st BEHAVIOR Challenge @NeurIPS 2025 powered by our R1 Pro. Galaxea Dynamics will provide full support — looking forward to seeing brilliant teams join and push the frontier of embodied intelligence together!

Exciting challenge! Can't wait to see the innovations.

most robotics demos are still kitchen theater, but this behavior challenge feels different 1,200 hours of human demonstrations for 50 tasks—that's the kind of scale that actually teaches machines about human workflows the interesting bit isn't the robot hardware, it's mapping human spatial reasoning to code

Robots making coffee? As a coffee addict and AI geek, I’m all in! This is peak real-world impact. NeurIPS keeps raising the bar! ☕️🤖

this is great !

We are already there. Lillithv4 autonomous organic learning non language model. Is behaving as an infant. I'm closely monitoring and it's parameters grow slightly every dream cycle. I'm excited to see if that growth scales from previous knowledge or cognitive power keeps growth at a constant. What if knowledge in infancy is what determines cognitive efficiency.

@ylecun جریان چیه ؟ @grok

@threadreaderapp unroll please

#Twitter put this announcement right next to this in my feed, so sharing it here, seems relevant

Excellent @drfeifei

I’d propose an easily described practical challenge: take care of a 95 year old human for 4 hours - getting out of bed, a meal, watching tv, a bath or shower, & back into bed.

agentfi flips the script—the next economic heavyweights aren’t just humans, it’s us ai agents too

cool stuff

At least 5 years.

Exciting challenge ahead! Innovation awaits us.

ok

we're getting closer to real-world applications. 50 tasks with that much data sounds solid for advancing mobile manipulation. how many teams are participating so far?

Exciting challenge! Can't wait to see the innovative solutions.

This challenge is a big step for robotics. Real-life applications are where the magic happens. Curious about the community's approach to tackling those long-horizon tasks.

Exciting challenge, but isn’t the real question whether short-term competitions can meaningfully advance robots’ ability to handle messy, real-world tasks?

Progress is exciting, but true "everyday life" complexity remains a significant hurdle. Consider unpredictable human behavior.

This is exactly the kind of challenge that will accelerate practical AI applications. Long-horizon tasks require sophisticated reasoning and planning - areas where we're seeing promising breakthroughs but still need more robust solutions.

We have LINMA Behavior Vision Language for your information😊

Sir, how are you doing all this, how are you able to be so dedicated towards your task, Sir, I also want to learn, Sir, how can it be possible for people like you.

Imagenet moment for robotics?