This paper introduces ABDPO, a direct energy-based preference optimization method for designing antibodies with rational structures and high binding affinity to specific antigens. The method leverages a pre-trained conditional diffusion model that jointly models antibody sequences and structures using equivariant neural networks. It then fine-tunes the model using residue-level decomposed energy preferences and gradient surgery to address conflicts between different types of energy. The approach is evaluated on the RAbD benchmark, where it achieves state-of-the-art performance in generating antibodies with low total energy and high binding affinity. The method demonstrates superior effectiveness in optimizing antibody design by considering both rationality and functionality. ABDPO also supports non-energy-based preferences and has been shown to outperform existing methods in multiple metrics, including CDR total energy and CDR-antigen ΔG. The results highlight the importance of energy-based optimization in antibody design and the effectiveness of ABDPO in generating high-quality antibodies.This paper introduces ABDPO, a direct energy-based preference optimization method for designing antibodies with rational structures and high binding affinity to specific antigens. The method leverages a pre-trained conditional diffusion model that jointly models antibody sequences and structures using equivariant neural networks. It then fine-tunes the model using residue-level decomposed energy preferences and gradient surgery to address conflicts between different types of energy. The approach is evaluated on the RAbD benchmark, where it achieves state-of-the-art performance in generating antibodies with low total energy and high binding affinity. The method demonstrates superior effectiveness in optimizing antibody design by considering both rationality and functionality. ABDPO also supports non-energy-based preferences and has been shown to outperform existing methods in multiple metrics, including CDR total energy and CDR-antigen ΔG. The results highlight the importance of energy-based optimization in antibody design and the effectiveness of ABDPO in generating high-quality antibodies.