paper/arxiv/2609.03276

Robotics Dataset

This paper introduces R2S-Eval, an evaluation pipeline for robot manipulation policies that combines real-to-sim calibration with vision-language model (VLM) preference evaluation to generate and assess rollout videos, producing reliable policy rankings without extensive hardware trials.

Catalog metadata

Repository owner

Yidi Wang, Feixiang Ruan, Ruoqu Chen et al.

Source

Research paper

Task

R2S-Eval: Robot Evaluation with Real-to-Sim Calibration via Vision-Language Models

Cameras

0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.