Tsinghua's EMERGE-Policy Orchestrates VLA, World Models as Multi-Agent Tools, Hits 94% Under Real-World Disturbances
机器之心 · wechat · 2026-09-12
Tsinghua SIGS and collaborators released EMERGE-Policy, an open-source multi-agent framework for robot manipulation (paper on arXiv). Instead of a single end-to-end VLA, it wraps VLA (pi0.5), motion primitives, world models (CosmosPolicy) and verifiers into a unified skill library orchestrated by graph-structured agents, letting system-level intelligence emerge from component collaboration.
Architecture
- A Main Agent decomposes long-horizon tasks and schedules via hierarchical context engineering; Perception/Verification/Monitor sub-agents run in isolated contexts.
- Three file-level memories: PLAN.md (task graph), HISTORY.md (observations), MEMORY.md (environment facts), auto-compressed when context overflows.
- The world model serves as a simulator to rehearse candidate action paths before execution, scored by verifiers — "think before acting."
Results: 99.2% on standard LIBERO (+0.7% over base); 93.9% under LIBERO-Plus perturbations (+11.7% over the CosmosPolicy base); strong RoboDojo-Sim memory scores. On a real-robot multi-layer cup-stacking task, it sustains 85%-94% success under human sabotage and cup/visual changes.
More from Embodied
- Agility Robotics lost $138.1M in 2025 as Brett Adcock's doom prediction nears — Rewkang · 2026-10-31
- Promised the Year of Physical AI: Humanoid Robots Are Behind Schedule — Distinct-Question-16 · 2026-10-31
- PokeBot's RoboAPI brings the call-a-model pattern to robot arms, no task training needed — rohanpaul_ai · 2026-10-31
- Foundry Robotics raises $100M Series A to build factories that look like data centers — ishabytes · 2026-10-31
- NVIDIA's MotionBricks robotics motion-generation model trends on Hugging Face — nvidia · 2026-10-31
- Agility Robotics to raise $465M ahead of Nasdaq debut as AGLT on Nov 5, Foxconn leads with $201M — xmercury_one · 2026-10-31