1 paper
Jierui Peng, Yanyan Zhang, Yicheng Duan +3
The evaluation of Vision-Language-Action (VLA) agents is hindered by the coarse, end-task success metric that fails to provide precise skill diagnosis or measure robustness to real…