StellarEdge Auriga report: verifiable execution harness hits 57.74% on 42 RoboDojo tasks
jiqizhixin · x · 2026-11-18
StellarEdge AI's Auriga technical report tackles robot recovery from failed grasps without resetting the scene.
- Bottleneck: understanding a task doesn't guarantee precise physical execution; a slip or collision changes the state the next action must start from.
- Mechanism: calibrated geometry (triangulation, VGGT-Omega depth, SAM 3.1) supplies target measurements; the model previews candidate trajectories while the world is paused; a single execution gateway checks tolerances, stops on divergence and returns observations for recovery; persistent task beliefs and evidence records let the model revise plans while keeping completed progress.
- Results: paired with GPT-6 Astra, Auriga averages 57.74% success across 42 RoboDojo tasks over 6,300 episodes; the frozen harness alone reaches 73.2%.
More from Embodied
- RoboDojo sim "not even wrong": blog exposes physics flaws in popular robot benchmark — ZeYanjie · 2026-11-18
- NVIDIA-backed Agile ONE SN humanoid robot for researchers ships early 2027 — xmercury_one · 2026-11-18
- Peking University Open-Sources PanDWM, a 360° 4D Driving World Model Beating DiST-4D — PekingUniversity · 2026-11-18
- Humanoids serve ice cream and spar in Shenzhen while APEC leaders visit — TansuYegen · 2026-11-18
- Otto by Rockwell explains using Gazebo digital twins for AMR deployments — ___Mufasaa · 2026-11-18
- Tokkey debuts thumb-sized AI hardware to run local models on your Mac — TejasKumar_ · 2026-11-18