Towards Rationale-Answer Alignment of LVLMs via Self-Rationale Calibration | Read Paper on Bytez