State-of-the-art cross-lingual transfer often relies on massive multilingual models, but their prohibitive size and computational cost limit their practicality for low-resource languages. An alternative is to adapt powerful, task-specialized monolingual models, but this presents challenges in bridging the vocabulary and structural gaps between languages. To address this, we propose KDA, a Knowledge Distillation Adapter framework that efficiently adapts a fine-tuned, high-resource monolingual model to a low-resource target language. KDA utilizes knowledge distillation to transfer the source model’s task-solving capabilities to the target language in a parameter-efficient manner. In addition, we introduce a novel adapter architecture that integrates source-language token embeddings while learning new positional embeddings, directly mitigating cross-lingual representational mismatches. Our empirical results on zero-shot transfer for Vietnamese Sentiment Analysis demonstrate that KDA significantly outperforms existing methods, offering a new, effective, and computationally efficient pathway for cross-lingual transfer.

Automatic microorganism detection, segmentation, and identification are essential to speeding up research in parasitology, biological treatment processes, and environment quality evaluation. Traditional methods based on microscopic images aim to find the differences in morphological features among the target species, such as outer shape and local features. This research focuses on species that share similar round shapes and seriously affect human health. Previous works show that segmentation is an essential step in improving detection accuracy. However, preparing segmentation ground truth is labor-intensive, especially in large datasets. For the dataset with no segmentation ground truths, we first generate pseudo-segmentation ground truths for training by applying a pre-trained segmentation network on a smaller dataset. We propose Attention-driven RetinaNet to detect, segment, and identify microorganisms in the microscopic images even when the training datasets have no annotated segmentation. The attention mechanism is applied to refine the incorrect pseudo-segmentation ground truths via Guided-attention. Self-attention is applied to select the essential part of the microorganism to improve detection performance. Experiments on the IEEE Parasitic Egg Detection and Classification in Microscopic Images Competition dataset show that the proposed method achieves 0.82 in mAP and outperforms other object detection methods.

Answerability on Visual Question Answering is a novel and attractive task to predict answerable scores between images and questions in multi-modal data. Existing works often utilize a binary mapping from visual question answering systems into Answerability. It does not reflect the essence of this problem. Together with our consideration of Answerability in a regression task, we propose VT-Transformer, which exploits visual and textual features through Transformer architecture. Experimental results on VizWiz 2020 dataset show the effectiveness and robustness of VT-Transformer for Answerability on Visual Question Answering when comparing with competitive baselines.