As of Oct 8, Robot Era says its VPP2 world action model ranks first on the RoboDojo simulation benchmark with a 32.26% average success rate and 39.26 average score, versus 22.48% and 28.97 for GPT-6-Astra, using only the standard dataset, QbitAI reports.
Robot Era says its VPP2 world action model ranks first on the RoboDojo simulation benchmark with a 32.26% average success rate and 39.26 average score, versus 22.48% and 28.97 for GPT-6-Astra, using only the standard dataset, QbitAI reports. In zero-shot tests on a real ALOHA dual-arm robot across 10 task types, VPP2 averaged 58.5% success against 40% for pi0.5 and led in 9 of 10. The model builds on Alibaba's open Wan2.1-I2V-14B, trains video prediction and action learning in separate stages, predicts 8 seconds of video in about 0.12 seconds after consistency distillation, and uses a 0.9B Action DiT.
Items older than 72 hours never go in Top stories and show their original date.