A tool to screen new ArXiv papers (github.com)

A new tool called **arxiv-digest** has been launched to assist researchers in efficiently sifting through the daily influx of technical papers on arXiv. Users can specify their research focus in a configuration file, incorporating positive and negative keywords, preferred authors, and optional instructions for a language model (LLM). The tool retrieves recent papers, assesses their relevance through a two-stage scoring system, and compiles a local HTML report highlighting the most significant findings. This is particularly useful for interdisciplinary topics that span multiple categories, such as "reinforcement learning in robotics," where pertinent papers may be distributed across various arXiv sections. The significance of arxiv-digest lies in its ability to streamline the research process, saving valuable time for academics and practitioners in the AI/ML community. The tool leverages both deterministic keyword scoring and LLM evaluation to provide a nuanced relevance score, enhancing the selection of papers based on specified academic interests. The user-defined keywords and scoring weights ensure that only the most pertinent research is highlighted, potentially leading to more informed and insightful contributions in rapidly evolving fields like AI and robotics. This innovative approach could pave the way for more personalized and effective literature review methods in various scientific domains.