The field of Natural Language Processing (NLP) is rapidly advancing in its ability to analyze and understand scientific literature. Recent developments have focused on improving the efficiency and accuracy of literature review generation, paper evaluation, and information retrieval. One notable trend is the use of large language models (LLMs) to automate tasks such as literature review generation, paper summarization, and question answering. These models have shown great promise in reducing the time and effort required to analyze and understand large volumes of scientific literature. However, challenges remain in ensuring the reliability and trustworthiness of these models, particularly in high-risk domains. To address this, researchers are exploring new methods for uncertainty quantification and conformal prediction, which can provide provable coverage guarantees and enhance the trustworthiness of LLMs. Noteworthy papers in this area include: GLiDRE, which introduces a new model for document-level relation extraction that achieves state-of-the-art performance in few-shot scenarios. Taggus, which proposes a pipeline for extracting social networks from literary fiction works in Portuguese, achieving satisfying results with an average F1-Score of 94.1% in character identification and 75.9% in interaction detection. Characterizing Deep Research, which proposes a formal characterization of the deep research task and introduces a benchmark to evaluate the performance of deep research systems. PaperEval, which presents a novel LLM-based framework for automated paper evaluation that addresses limitations in outdated domain knowledge and limited reasoning capabilities. Conformal Sets, which proposes a frequency-based uncertainty quantification method under black-box settings, leveraging conformal prediction to ensure provable coverage guarantees.