WorldOracle world model evaluation platform released, using real-machine testing to fact-check the industry
On October 16, WorldOracle, an action-driven world model evaluation platform jointly completed by researchers from AgiBot, NVIDIA, Princeton University, the University of Texas at Austin, and the University of Toronto, was officially released and opened evaluation submissions to developers.
Highlights
X Square Robot releases X2Real simulation evaluation benchmark, with correlation between simulation and real-machine task success rates reaching 0.96
X Square Robot released the evolvable simulation evaluation benchmark X2Real, whose simulation evaluation and real-machine evaluation achieved a task success rate correlation of 0.96 and a progress score correlation of 0.99.
自变量机器人 · 微信公众号Source ↗Evaluation of a frozen frontier VLM used directly as a robot policy on the RoboMME memory-dependent manipulation benchmark
The study uses the frozen frontier vision-language model GPT-6-Astra directly as a robot policy and tests it on the RoboMME benchmark, which contains 16 memory-dependent manipulation tasks; the model retains the entire episode context throughout and directly outputs each end-effector waypoint and gripper command, without any training specific to this benchmark.
arXiv cs.ROSource ↗SenseTime launches SenseNova-RoboRSI and a 27B embodied model, achieving SOTA on 12 simulation benchmarks
The SenseTime SenseNova team launched the SenseNova-RoboRSI unified execution framework and the 27B embodied model SenseNova-RoboRSI-27B-Preview.
Xbotics具身智能实验室 · 微信公众号Source ↗
Model release/update
SenseTime launches SenseNova-RoboRSI and a 27B embodied model, achieving SOTA on 12 simulation benchmarks
The SenseTime SenseNova team launched the SenseNova-RoboRSI unified execution framework and the 27B embodied model SenseNova-RoboRSI-27B-Preview.
Xbotics具身智能实验室 · 微信公众号Source ↗BAAI Zhongzhi FlagOS 2.2 released: more than 1,454 operators, 27% cross-chip performance improvement
The BAAI (Beijing Academy of Artificial Intelligence) showcased FlagOS 2.2 progress at the 2026 AI Open Computing Conference and Zhongzhi FlagOS Technology Conference; this version was released on September 30. Public operators in the nine major operator libraries increased from 564 to 1,454 (890 newly added), and in tests covering over a thousand operators and 10 heterogeneous AI chips, cross-chip operator performance improved by about 27% over the previous version; FlagOS-Compressor, a weight conversion and quantization tool, was released for the first time, enabling a new model to adapt to as many as 12 chips within one day.
智源研究院 · 微信公众号Source ↗
Product release/update
SUPCON UCS put into production at Qinghai Salt Lake Magnesium Industry's 300,000-ton PVC plant, with TPT deployed simultaneously
SUPCON's self-developed universal control system UCS was successfully put into production in the intelligent transformation project of Qinghai Salt Lake Magnesium Industry's 300,000-ton ethylene-based PVC plant, becoming the first large-scale UCS production case in China's ethylene-based PVC industry; at the same time, the time-series large model TPT won the bid and entered the deployment and implementation stage.
Paper research
WorldOracle world model evaluation platform released, using real-machine testing to fact-check the industry
On October 16, WorldOracle, an action-driven world model evaluation platform jointly completed by researchers from AgiBot, NVIDIA, Princeton University, the University of Texas at Austin, and the University of Toronto, was officially released and opened evaluation submissions to developers.
X Square Robot releases X2Real simulation evaluation benchmark, with correlation between simulation and real-machine task success rates reaching 0.96
X Square Robot released the evolvable simulation evaluation benchmark X2Real, whose simulation evaluation and real-machine evaluation achieved a task success rate correlation of 0.96 and a progress score correlation of 0.99.
自变量机器人 · 微信公众号Source ↗Evaluation of a frozen frontier VLM used directly as a robot policy on the RoboMME memory-dependent manipulation benchmark
The study uses the frozen frontier vision-language model GPT-6-Astra directly as a robot policy and tests it on the RoboMME benchmark, which contains 16 memory-dependent manipulation tasks; the model retains the entire episode context throughout and directly outputs each end-effector waypoint and gripper command, without any training specific to this benchmark.
arXiv cs.ROSource ↗RAX: an open-source framework that expands robot agent capabilities through online exploration
The paper proposes RAX (Robot Agent eXploration), an open-source robot agent framework centered on expanding online exploration, in which a general coding agent generates robot control code without using action demonstration data, manually designed action primitives, or value function guidance.
arXiv cs.ROSource ↗