Tag
EXIMO proposes an efficient algorithm for fine-tuning vision-language-action robot policies using a three-stage process: VLM-guided exploration, imitation learning, and residual reinforcement learning, showing improved sample-efficiency and performance.
A tweet highlighting that benchmarking robot policies is broken and sharing results from thousands of evaluations over 12 manipulation tasks to determine which policy to use.
RoboTTT scales visuomotor context to 8K timesteps for robot policies, enabling one-shot imitation from human video demonstrations, on-the-fly policy improvement, and robustness to perturbations. It achieves an 87% improvement over baselines and completes a five-minute, ten-stage assembly task that no baseline could.
Fei-Fei Li highlights a new test-time training approach for robotic learning, developed in collaboration between Stanford SVL and NVIDIA Robotics, which scales robot model context to 8000 timesteps with constant inference cost.
Evaluation of 30+ embodied AI models finds that current generalist robot policies lack robustness for real-world manipulation, leading to the creation of RoboDojo.
RoboTALES introduces a two-stage framework combining LLM-based planning and VLM-based criticism to improve task-aligned video generation and robotic policy training, significantly outperforming existing methods on long-horizon manipulation tasks.
This paper introduces In-Context World Modeling (ICWM), a framework that enables robot policies to infer system variables from self-generated interactions, allowing adaptation to novel configurations without parameter updates by treating system identification as an in-context adaptation problem. It outperforms standard VLA baselines on novel camera viewpoints in simulation and real-world experiments.
Sergey Levine announces he will be speaking at CVPR workshops on test-time scaling for computer vision and robot policy generalization, as well as on deployment of foundation models.
AURA-Mem proposes a constant-size memory for robot policies using a learned gate that writes only when current observations would change the next action. It matches baseline accuracy with significantly fewer writes and constant VRAM, addressing the memory bottleneck for long-horizon robot tasks.