“The objective of the Chinese Vision-Language Understanding Evaluation (CVLUE) is to comprehensively assess the performance of Chinese vision-language multimodal pre-trained models in multimodal modeling and understanding across four tasks: Image-Text Retrieval, Visual Question Answering, Visual Grounding, and Visual Dialog. To enhance the models’ performance across various multimodal tasks, this paper propose a multimodal information understanding enhancement method based on answer-guided images. Firstly, we propose task-specific methods for answer-guided image generation. Secondly, the authentic and answer-guided images are fed into the model for multimodal fine-tuning, respectively. Finally, training objectives are set for different tasks to minimize the gap between the answer-guided images and authentic images, thereby supervising the results produced by the authentic images utlizing answer-guided images. The experimental results demonstrate the effectiveness of the proposed method.”

Multimodal machine translation (MMT) simultaneously takes the source sentence and a relevant image as input for translation. Since there is no paired image available for the input sentence in most cases, recent studies suggest utilizing powerful text-to-image generation models to provide image inputs. Nevertheless, synthetic images generated by these models often follow different distributions compared to authentic images. Consequently, using authentic images for training and synthetic images for inference can introduce a distribution shift, resulting in performance degradation during inference. To tackle this challenge, in this paper, we feed synthetic and authentic images to the MMT model, respectively. Then we minimize the gap between the synthetic and authentic images by drawing close the input image representations of the Transformer Encoder and the output distributions of the Transformer Decoder. Therefore, we mitigate the distribution disparity introduced by the synthetic images during inference, thereby freeing the authentic images from the inference process. Experimental results show that our approach achieves state-of-the-art performance on the Multi30K En-De and En-Fr datasets, while remaining independent of authentic images during inference.