Humanoid robots are increasingly deployed to perform tasks involving dexterous manipulation and heavy-load carrying. However, when handling objects with an offset center of mass, uneven force distribution between the arms and reduced dynamic stability can arise, leading to higher energy consumption, motor overheating, or degraded control performance. A promising solution is to use coordinated dual-arm manipulation to adjust the object’s pose, thereby reducing the negative effects of load imbalance. Achieving this requires precise coordination of motion and force between both arms. To this end, we propose a collaborative pose optimization strategy for dual-arm systems that explicitly incorporates object inertial properties into the planning process. The pose adjustment is formulated as a constrained optimization problem considering kinematic feasibility, grasp equilibrium, torque limits, and collision avoidance. The resulting optimal terminal pose serves as a reference for trajectory planning. We further employ an MPC-based planner to generate joint-space motions that track the adjustment trajectory while satisfying task constraints. Simulation results show that the proposed method reduces joint torque under various eccentric loading conditions while ensuring stable, constraint-compliant execution through active object pose adjustment. The resulting motions satisfy joint limits, torque constraints, and collision avoidance, achieving a balanced and robust manipulation configuration.