(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
41
276
1,432
456,027
(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: behavior.stanford.edu
2
3
52
9,805
(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: arxiv.org/abs/2403.09227
What’s new in this challenge?
2
7
58
9,153
(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
Sep 2, 2025 · 8:12 PM UTC
3
5
44
7,804
(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
2
7
38
6,690
(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
2
3
32
5,929
(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
2
4
28
5,168
(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
2
2
25
4,487
(9/N) 🧪Evaluation & Submission
Submission instructions and evaluation details are available on our website: behavior.stanford.edu/challe…
Start experimenting today and get ready to compete!
Deadline: Nov. 15th
Winners announced: Dec. 1st
NeurIPS challenge: Dec. 6-7, San Diego, CA
1
2
28
4,523
(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?
2
1
25
4,150
(11/N) 💬Join our Discord to ask questions and discuss:
🔗 discord.gg/bccR5vGFEx
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.
2
1
25
4,096
(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
…
1
2
25
3,560
(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
1
2
21
3,562
(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.
5
5
35
16,563
(N/N) We thank our sponsors for their generous support:
@SimovationInc
@IMDAsg
@StanfordHAI
@SchmidtFutures
6
4
37
15,065
