Enhancing Cross-Modal Fine-Tuning with Gradually Intermediate Modality Generation

Lincan Cai, Shuang Li, Wenxuan Ma +4

Large-scale pretrained models have proven immensely valuable in handling data-intensive modalities like text and image. However, fine-tuning these models for certain specialized mo…