BEHAVIOR Challenge

How to read this picture

Annual simulation competition on 50 (2025) or 100 (2026) BEHAVIOR-1K household tasks with a bimanual mobile robot.1

Sources
Last checked 10 Oct 2026Basic entry15 of 20 facts checked at the sourceNext check 8 Apr 2027
Runs in
Simulation1
Checked against real robots
Not checked
Skill
Household tasks
Robot
Arm on wheels, Two arms2
Licence
Unclear3

Comparisons with real robots

No comparison found Unknown

Details

No real-robot evaluation of challenge entries found on the challenge pages, the HAI announcement, or the top-2 team reports. The only BEHAVIOR sim-to-real data is the one-task study in the BEHAVIOR-1K paper (see that record).

Details

About

What it is
Competition Inferred4
More

Classified by the Atlas from how the authors describe and distribute it.

Built by
Stanford Vision and Learning Lab (page footer); 2025 sponsors: Simovation, IMDA, Stanford HAI, Schmidt Family Foundation, NVIDIA; 2026 sponsors: Simovation, IMDA, Stanford HAI, Schmidt Family Foundation5
More

2026 sponsors from https://behavior.stanford.edu/challenge/index.html

Released
2025-09 (2025 challenge launched 2025-09-02; the 2026 page calls 2026 the 'second year')5
More

A predecessor 'BEHAVIOR Challenge 2021' on BEHAVIOR-100 (100 activities, iGibson) ran on EvalAI 2021-07-15 to 2022-04-01: https://eval.ai/web/challenges/challenge-page/1190/overview

Version
2026 BEHAVIOR Challenge, 100 tasks, single track (RGB + depth + proprioception)6
Last update
2026-10-09: final deadline 2026-10-16 AoE; new rules on evaluation speed, timeouts, reconnections, single-run submissions. 2026-10-07: move to v3.9.3-post2.7
Status
Active Inferred7
More

2026 edition open; updates posted 2026-10-09.

Setup

Runs in
Simulation1
Robot
Arm on wheels, Two arms2
More

Challenge use from https://behavior.stanford.edu/challenge/archive/2025/evaluation.html and https://behavior.stanford.edu/challenge/evaluation.html. The official pages do not name the manufacturer.

Setting
Whole home1

Scoring and access

Scored by
Progress score, Success rate6
Trials
2025: participants self-evaluate on 10 instances per task, 1 rollout each; organisers evaluate top 5 on 10 held-out instances per task. 2026: 20 public + 20 hidden instances per task; reported results on instances 0-9, 1 rollout each; top-5 re-run on hidden instances.6
More

2025 from https://behavior.stanford.edu/challenge/archive/2025/evaluation.html. The 2026 page describes the hidden set in two slightly different ways.

Who runs it
Both teams and organisers8
Leaderboard
Official leaderboard8
More

2026: https://huggingface.co/spaces/behavior-1k/2026-challenge-leaderboard

Code licence
MIT (BEHAVIOR-1K repository LICENSE)3
Data licence
Demos: MIT (HF cards behavior-1k/2025-challenge-demos, behavior-1k/2026-challenge-demos). Assets: BEHAVIOR Data Bundle EULA, non-commercial academic research only.9
More

Asset EULA from https://github.com/StanfordVL/BEHAVIOR-1K/blob/main/setup.sh; demos card also https://huggingface.co/datasets/behavior-1k/2025-challenge-demos

Sources 9

  1. 1馃弳 2026 BEHAVIOR Challenge - BEHAVIOROfficial site 路 checked 10 Oct 2026
  2. 2Robots - BEHAVIOROfficial site 路 checked 10 Oct 2026
  3. 3StanfordVL/BEHAVIOR-1K on GitHub (blob)Repository 路 checked 10 Oct 2026
  4. 4BEHAVIOR-1K: A Human-Centered, Embodied AI Benchmark with 1,000 Everyday Activities and Realistic SimulationPaper 路 Mar 2024 路 checked 10 Oct 2026
  5. 5馃弳 2025 BEHAVIOR Challenge - BEHAVIOROfficial site 路 checked 10 Oct 2026
  6. 6Evaluation and Rules - BEHAVIOROfficial site 路 checked 10 Oct 2026
  7. 7Announcements / Updates - BEHAVIOROfficial site 路 checked 10 Oct 2026
  8. 8Challenge Leaderboard - BEHAVIORLeaderboard 路 checked 10 Oct 2026
  9. 9behavior-1k/2026-challenge-demos on Hugging Face (dataset)Repository 路 checked 10 Oct 2026

Change history

  1. Created as a basic entry: identity facts checked at primary sources (phase 1 re-verification).