WAM is capable in BEHAVIOR-1K: the 2nd-place solution to the 2026 BEHAVIOR Challenge
WAM is capable in BEHAVIOR-1K: the 2nd-place solution to the 2026 BEHAVIOR Challenge
This work presents the second-place solution (Team IF1001) to the 2026 BEHAVIOR Challenge, which evaluates mobile manipulation across 100 household tasks.
This submission reports the second-place solution to the 2026 BEHAVIOR Challenge, achieving a mean Q-score of 0.615 and a full-task success rate of 41.2% on an official private test set with 20 instances per task across 100 household tasks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported success rate
40.9%
Reported trials
View original evidence
o closed-source models; during official private-test evaluation, our system handled concurrent requests across 100 evaluation ports with limited hardware resources, sustaining an average end-to-end inference rate above 2 FPS per port. Our method achieves a mean Q-score of 0.613 and a full-task success rate of 40.9%, averaged over 3 self-evaluation rounds across all 100 tasks. On the official private test set, with 20
Open source S4Reported success rate
41.2%
Reported trials
View original evidence
instances per task, our system achieves a mean Q-score of 0.615 and a full-task success rate of 41.2%.
Open source S5mean Q-score
0.613
View original evidence
o closed-source models; during official private-test evaluation, our system handled concurrent requests across 100 evaluation ports with limited hardware resources, sustaining an average end-to-end inference rate above 2 FPS per port. Our method achieves a mean Q-score of 0.613 and a full-task success rate of 40.9%, averaged over 3 self-evaluation rounds across all 100 tasks. On the official private test set, with 20
Open source S4mean Q-score
0.615
View original evidence
instances per task, our system achieves a mean Q-score of 0.615 and a full-task success rate of 41.2%.
Open source S5inference rate
2
View original evidence
o closed-source models; during official private-test evaluation, our system handled concurrent requests across 100 evaluation ports with limited hardware resources, sustaining an average end-to-end inference rate above 2 FPS per port. Our method achieves a mean Q-score of 0.613 and a full-task success rate of 40.9%, averaged over 3 self-evaluation rounds across all 100 tasks. On the official private test set, with 20
Open source S4evaluation ports
100
View original evidence
o closed-source models; during official private-test evaluation, our system handled concurrent requests across 100 evaluation ports with limited hardware resources, sustaining an average end-to-end inference rate above 2 FPS per port. Our method achieves a mean Q-score of 0.613 and a full-task success rate of 40.9%, averaged over 3 self-evaluation rounds across all 100 tasks. On the official private test set, with 20
Open source S4household tasks
100
View original evidence
arXiv:2611.13818v1 Announce Type: new Abstract: This work presents the 2nd-place solution (Team IF1001) to the 2026 BEHAVIOR Challenge, which evaluates mobile manipulation across 100 household tasks. Our method is built on a powerful world action model (WAM), Cosmos 3. We improve Cosmos 3 by using richer visual history, task-progress conditioning, and task-group and skill-specific LoRA adapters. During inference, we
Open source S2
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
o closed-source models; during official private-test evaluation, our system handled concurrent requests across 100 evaluation ports with limited hardware resources, sustaining an average end-to-end inference rate above 2 FPS per port. Our method achieves a mean Q-score of 0.613 and a full-task success rate of 40.9%, averaged over 3 self-evaluation rounds across all 100 tasks. On the official private test set, with 20
Open source S4
instances per task, our system achieves a mean Q-score of 0.615 and a full-task success rate of 41.2%.
Open source S5
arXiv:2611.13818v1 Announce Type: new Abstract: This work presents the 2nd-place solution (Team IF1001) to the 2026 BEHAVIOR Challenge, which evaluates mobile manipulation across 100 household tasks. Our method is built on a powerful world action model (WAM), Cosmos 3. We improve Cosmos 3 by using richer visual history, task-progress conditioning, and task-group and skill-specific LoRA adapters. During inference, we
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org