Skip to content

πŸ† 2025 BEHAVIOR Challenge

Join us and solve 50 full-length household tasks in the realistic BEHAVIOR-1K environment, with 10,000 teleoperated expert demonstrations (1200+ hours) available! πŸ€–

πŸ“£ Announcements

πŸ—“οΈ 11/07/2025 β€” Rule Clarifications

Key updates this week:

  • Documentation updates
  • Bug fixes & Hidden test logic addition.

Read full details β†’

πŸ—“οΈ 10/30/2025 β€” Rule Clarifications & Features Update

Key updates this week:

  • We have released language annotations for all 50 tasks.
  • We have included task id as part of the observation dict that the policy will receive.
  • Added more utilities (HeavyRobotWrapper, score_utils, etc.)

Read full details β†’

πŸ—“οΈ 10/08/2025 β€” Rule Clarifications, Bug Fixes & NVIDIA Sponsorship

Key updates this week:

  • Clarified evaluation setup: only task-relevant object poses and the robot’s initial pose will be randomized.
  • Privileged info allowed during training for both tracks.
  • Multiple bug fixes (eval_utils.py, USD asset format, partial credit).
  • Updated submission guidelines + sample Dockerfile.
  • New sponsor: NVIDIA!
  • πŸ’° Prize pool updated:
    • 1st: $1,000 + GeForce 5080
    • 2nd: $500 + (Jetson Orin Nano Super or $1,000 Brev Credits)
    • 3rd: $300 + $500 Brev Credits

Read full details β†’

πŸ—“οΈ 09/28/2025 β€” Dataset Fixes & CLI Improvements

Highlights:

  • No formal registration required β€” submit directly!
  • Fixed dataset sharding, robot start poses, and improved baseline checkpoints.
  • Added new CLI args for evaluation (testing_on_train_instances, max_steps, partial_scene_load).

Read full details β†’

πŸ—“οΈ 09/19/2025 β€” Rule Clarifications, Evaluation Protocol & Tutorial

Highlights:

  • BDDL task definitions can be used for both tracks and are identical during evaluation.
  • Additional self-collected data allowed for both tracks.
  • Defined evaluation timeout and success score metrics.
  • Various bug fixes (Windows setup, dataset timestamp, evaluation scripts).
  • Added new tutorial: Configure robot action space.

Read full details β†’

Overview

BEHAVIOR is a robotics challenge for everyday household tasks. It's a large-scale, human-grounded benchmark that tests a robot's capability in high-level reasoning, long-range locomotion, and dexterous bimanual manipulation in house-scale scenes.

This year's challenge features:

  • 50 full-length household tasks from our 1,000 activity collection, covering diverse activities like rearrangement, cooking, cleaning, and installation
  • 10,000 teleoperated demonstrations (1200+ hours) for training

BEHAVIOR challenge is co-hosted with the Embodied Agent Interface Competition at NeurIPS 2025.

Dataset & Baselines

Teleoperated Demonstrations

10,000 expert demonstrations (1200+ hours) collected via teleoperation:

  • Synchronized RGBD observations
  • Object and part-level segmentation
  • Ground-truth object states
  • Robot proprioception and actions
  • Skill and subtask annotations

Dataset details β†’

Baseline Methods

Pre-implemented training & evaluation pipelines for:

  • Behavioral Cloning baselines: ACT, Diffusion Policy, BC-RNN, WB-VIMA - these are diverse imitation learning approaches that learn from the provided demonstrations.
  • Pre-trained Visuo-Language Action models: OpenVLA and Ο€0. These models are pretrained by a large amount of demonstration data, giving an alternative to models that need to be trained from scratch.

Baselines details β†’

Evaluation & Rules

The organizers reserve the right of final interpretation of the challenge rules.

Challenge Tracks

Standard track: Limited to provided robot onboard observations (RGB + depth + instance segmentation + proprioception).

Privileged information track: May query simulator for any information (object poses, scene point clouds, etc.).

πŸ† Prizes per track:

  1. πŸ₯‡ $1,000 + GeForce 5080
  2. πŸ₯ˆ $500 + (Jetson Orin Nano Super or $1,000 Brev Credits)
  3. πŸ₯‰ $300 + $500 Brev Credits

Top 3 teams from each track will be invited to present at the workshop!

Evaluation Metrics

Primary metric (for ranking): Task success rate averaged across 50 tasks. Partial credit given as fraction of satisfied BDDL goal predicates.

Secondary metrics (efficiency):

  • Simulated time - Total simulation steps Γ— time per step
  • Distance navigated - Total base movement distance
  • Hand displacement - Cumulative hand movement

Evaluation details & Full challenge rules β†’

Participating

Resources

Join our community to ask questions and discuss the challenge:

Whether you're a robotics veteran or just entering the field, we're here to support you.

Important Dates

  • Challenge Launch: September 2, 2025
  • Submission Deadline: November 15, 2025
  • Winners Announcement: December 6-7, 2025 @ NeurIPS conference in San Diego

BibTeX

To cite BEHAVIOR-1K, please use: 

@article{li2024behavior,
  title={Behavior-1k: A human-centered, embodied ai benchmark with 1,000 everyday activities and realistic simulation},
  author={Li, Chengshu and Zhang, Ruohan and Wong, Josiah and Gokmen, Cem and Srivastava, Sanjana and Mart{\'i}n-Mart{\'i}n, Roberto and Wang, Chen and Levine, Gabrael and Ai, Wensi and Martinez, Benjamin and Yin, Hang and Lingelbach, Michael and Hwang, Minjune and Hiranaka, Ayano and Garlanka, Sujay and Aydin, Arman and Lee, Sharon and Sun, Jiankai and Anvari, Mona and Sharma, Manasi and Bansal, Dhruva and Hunter, Samuel and Kim, Kyu-Young and Lou, Alan and Matthews, Caleb R. and Villa-Renteria, Ivan and Tang, Jerry Huayang and Tang, Claire and Xia, Fei and Li, Yunzhu and Savarese, Silvio and Gweon, Hyowon and Liu, C. Karen and Wu, Jiajun and Fei-Fei, Li},
  journal={arXiv preprint arXiv:2403.09227},
  year={2024}
}

Sponsors

High-quality simulation data provided by Simovation.

We gratefully acknowledge the support of our sponsors who make this challenge possible: