ShengShu Technology Unveils Motus2 for Robotic Dexterous Manipulation

SINGAPORE, Sept. 14, 2026 /CourierPR/ -- ShengShu Technology, a company founded in 2023, has unveiled Motus2, a groundbreaking self-evolving general world model designed for robotic dexterous manipulation. The technology, unveiled at the 2026 Inclusion Conference on the Bund on September 10, aims to enhance the capabilities of robots in performing complex tasks through a unified model that integrates action generation, consequence prediction, and outcome evaluation.

Motus2 operates by generating candidate actions based on language instructions, robot states, and visual history. A simulator interface predicts the future visual states these actions might produce, while an evaluator interface assesses whether the predicted outcomes would advance the task. Together, these components form a closed loop that continuously refines the robot's actions. During execution, the model uses Best-of-N planning to generate multiple candidate actions, predict and compare their outcomes, and execute the highest-scoring option. After each action chunk, the robot incorporates fresh observations and plans its next step. In training, the model updates its policy based on value scores assigned to candidate actions, making it more likely to generate actions that support task completion.

The company's research paper reports that Motus2 achieved an average success rate of 84% across five primary real-robot tasks: placing a ball, multi-finger manipulation, attaching an eraser, screwing in a light bulb, and placing a phone. In a separate policy-optimization study involving phone placement and multi-finger manipulation, combining model-based reinforcement learning with inference-time planning increased the average success rate from 65% to 75%, a gain of 10 percentage points. These evaluations were based on 20 trials per task for each method.

ShengShu Technology's Motus2 addresses the challenges of dexterous robotic manipulation, which require coordinated handling of spatial relationships, finger movements, sustained contact, and historical information. The model has been evaluated on dual-arm robot platforms equipped with WUJI, WUJI Hand 2, and Sharpa Wave dexterous hands.

To enhance its performance, Motus2 incorporates mechanisms for retaining historical information and a lightweight tactile expert. The tactile expert reads the latest contact feedback immediately before a short action segment is executed and refines the action accordingly. Adding the tactile expert increased the average success rate for two real-robot tasks, pulling out a paper cup and tearing paper, from 60% to 72.5%, a gain of 12.5 percentage points.

The model architecture, research paper, and real-robot demonstrations are publicly available. ShengShu Technology, which focuses on the research and development of general world models, aims to build general intelligence systems capable of understanding, predicting, and acting in both digital and physical worlds.