Move-Then-Operate: Behavioral Phasing for Human-Like Robotic Manipulation

We present Move-Then-Operate, a Vision–language–action framework that explicitly decouples robotic manipulation into two distinct behavioral phases: coarse relocation (move) and contact-critical interaction (operate). Unlike monolithic policies that conflate these heterogeneous regimes, our architecture employs a dual-expert policy routed by a learnable phase selector, introducing a structural inductive bias that isolates phase-specific dynamics. Phase labels are automatically generated via an MLLM-based pipeline conditioned on lightweight contextual cues such as end-effector velocity and subtask decomposition to ensure alignment with human motor patterns. Evaluated on the RoboTwin2 benchmark, our method achieves an average success rate of $68.9\%$, outperforming the monolithic $\pi_0$ baseline by +$24\%$. It matches or exceeds models trained on $10\times$ more data and reaches peak performance in $40\%$ fewer training steps, demonstrating that architectural disentanglement of move and operate phases is a highly effective and efficient strategy for mastering high-precision manipulation.

Our paper aims to make robots better at everyday manipulation tasks by separating them into two natural stages: first moving close to the target, and then carefully interacting with it. Instead of learning the whole task as one mixed behavior, our method uses two specialized parts, one for moving and one for precise operation, with a small decision module choosing which part to use at each moment. The training labels for these stages are generated automatically, avoiding heavy manual annotation. Experiments on RoboTwin2 show that this simple “move then operate” strategy improves robot success rates to 68.9%, outperforming the baseline by 24%, while also requiring less training data and fewer training steps.