GPT-6-Astra evaluated on EBench for direct and policy-harness robot control
Direct control leads on short tasks but lags on long-horizon work; calling a VLA or WAM matches top policies at lower token cost.
Stories per day · 30 days
2 RSS
Relatedrobotics2embodied-ai1gpt-6-astra1libero-x1mllm1robot-control1splines1vela1vision-language-action1vla1
2 stories, newest first · one card per story
Stories tagged #ebench
Vela: Scaling Vision-Language-Action Models with Adaptive Action Curve Parametrization
Vela models robot actions as adaptive continuous trajectories and reports gains on simulation and real manipulation tasks.