Cortex aligns planning and execution for long-horizon robot tasks and hits 95.5% on LIBERO-Long
机器之心 · wechat · 2026-07-22
Cortex proposes a bidirectionally aligned embodied agent for long-horizon manipulation, linking a high-level planner and a low-level executor through structured subtask descriptions and memory.
The system splits work into System-2 planning and System-1 execution, standardizes 32 skills, and uses event-balanced sampling to focus training on task boundaries. In experiments, it reaches 95.5% on LIBERO-Long, 86.8% on RoboTwin2.0, and succeeds on real dual-arm chemistry workflows where end-to-end baselines fail mid-task.
The main takeaway is that long-horizon robot competence depends not just on action generation, but on maintaining task progress, state, and recovery across steps.
More from Embodied
- NVIDIA releases MotionBricks: real-time humanoid motion generation at 15,000 FPS with 2ms latency — _akhaliq · 2026-10-26
- Claude Code finds root backdoor in Honda Civic head unit, then debloats it — IndraVahan · 2026-10-26
- Temple health wearable opens pre-orders, users ask: why wear it before benefits are proven — abhishekcode42 · 2026-10-26
- FLASH: GPU-native deformable sim learns cloth folding in minutes on one RTX 5090 — ChongZzZhang · 2026-10-26
- Luke Wroblewski spots it: iPhone Duo's corner radius mirrors the original iPhone — LukeW · 2026-10-26
- Humanoids shattered forecasts: half-marathon record in 2026, 100m faster than Bolt — ben_j_todd · 2026-10-26