The field of natural language processing and multimodal models is witnessing significant developments, with a focus on improving the performance and equity of large reasoning models in multilingual settings. Researchers are exploring the internal reasoning processes of these models, revealing linguistic biases and highlighting the need for more inclusive and diverse training data. In the area of misinformation detection, novel frameworks and datasets are being proposed to tackle the challenges of out-of-context news and image-text claims. These innovations have the potential to enhance the accuracy and reliability of misinformation detection systems. Notable works in this area include the development of datasets such as AVerImaTeC, which provides annotated evidence for image-text claims, and the proposal of frameworks like CMIE, which combines multimodal large language model insights with external evidence for explainable out-of-context misinformation detection. Other significant contributions include the introduction of contrastive distillation methods for transferring emotional knowledge from large language models to smaller models, and the development of interactive web platforms for historical image age estimation. Some papers that are particularly noteworthy include:

• AVerImaTeC, which introduces a novel dataset for automatic verification of image-text claims with evidence from the web.
• CMIE, which proposes a framework that incorporates coexistence relationship generation and association scoring mechanisms to enhance misinformation detection.
• Contrastive Distillation of Emotion Knowledge, which presents a method for transferring rich emotional knowledge from large language models to compact models without human annotations.