For Corporate Bond Credit Rating (CPCR) prediction, instances in training datasets are assigned by one or more logical labels, indicating whether an instance belongs to a specific class. However, Label Distribution (LD), which reveals credit rating correlation information between labels, is usually ignored due to the difficulty of obtaining the label distributions directly. We present an easy-to-use method named FBLE to generate the label distributions from the logical labels in the training set via leveraging the relevance between labels and instances, being capable of improving the generalization of models without adding any extra parameters. Such a process of generating label distributions from logical labels is called Label Enhancement (LE), which reinforces the supervision information in the training sets. We constructed a CPCR dataset to show that models with FBLE could achieve state-of-the-art or competitive results. Electronic document is a “live” template and already defines the components of your paper [title, text, heads, etc.] in its style sheet.