This work introduces A2-MEAS, an embodied multi-agent framework built around code as a scaling interface that combines corrective language, spatial evidence, execution history, and action proposals to support efficient embodied learning.
Abstract
Foundation models increasingly guide robot actions, yet sustained autonomy requires efficient control and failure recovery. We introduce A2-MEAS, an embodied multi-agent framework built around code as a scaling interface. Learned correction codes combine corrective language, spatial evidence, execution history, and action proposals. MetaBrain provides training supervision and runtime guidance. A correction encoder and adapter absorb this knowledge into policy weights during training, while fresh codes adapt behavior with frozen weights at deployment. The action model, A2-WLA, adds motion-aware history and a latent world model to a vision-language-action policy. Surprisingly, this simple design with action pretraining and task specialization reaches 0.5755 Q-score on the 50-task BEHAVIOR 2026 public subset before online correction, 2.21x the 2025 winner's reported 0.2605 across different benchmark editions and training configurations. On all 100 BEHAVIOR 2026 public tasks, A2-MEAS achieves 0.6411 Q-score and 45.40% task success. On RoboTwin 2.0, A2-WLA achieves 93.98% average success, rising to 95.05% with the complete system. Across 12 real-robot tasks in five household settings, A2-WLA achieves 57.92% average success versus 26.67% for pi0.5. In long-horizon pour-over coffee preparation, a history-conditioned adapter raises post-failure completion at two filter-handling steps from 14% to 94% and from 8% to 82%, compared with A2-WLA retries. Our training infrastructure delivers 427.1 samples per second, 4.03x the throughput of a pi0.5 reference implementation on matched hardware. We will release the system, models, training scripts, and infrastructure to support efficient embodied learning.
EmbodiedEvo is a self-evolving coding agent that distills physical feedback into reusable code Skills and diagnostic Lessons that improves weaker executors and shows the largest gains on memory-dependent tasks and successful reuse of coordinated procedures.
OmniBrain-0 is introduced, a lightweight unified embodied agent that coordinates manipulation and navigation through a shared high-level decision model, integrating task understanding, planning, progress tracking, and feedback-driven recovery.
Xiaomi Embodied Intelligence Team Yuechen Luo, Chong He, Fang Li et al.· 2 citations
The results show that value models provide useful trajectory ranking and failure-detection signals in the studied manipulation setting and improve assisted task completion and reduce takeover frequency relative to recovery-only training, although these gains must be distinguished from fully autonomous policy performanc...
Hong-Liang Wang, Jin Liu, Xiao-Lin Zhu et al.· ITU Journal on Future and Ev...· 0 citations
FIND is introduced, an agentic real-world RL framework that closes the loop between scene understanding, weakness-aware practice, self-evaluation, and policy improvement in a persistent workspace and reframes autonomous practice as a scene-conditioned, performance-aware task-selection problem.
EmbodiedEvo is introduced, where a coding agent evolves reusable Skills and Lessons through feedback on a single Playground scene, then freezes the resulting harness for diverse new scenes, to retain and refine experience for later scenes.
AW-0, an agentic world action model framework that integrates code-based planning with joint video and action prediction, is introduced and an automated curation pipeline that labels robot and human video trajectories with aligned programs and consolidates recurring routines into reusable skill libraries is introduced.
Zhuo-Yang Liu, Dan-Tong Niu, Bo-Tong Zhang et al.· 2 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.