Automatic extraction of high-quality biomedical entity relations from biomedical texts plays an important role in biomedical text mining. Currently, existing methods generally focus on training a single task model for a specific task (