The field of scientific information retrieval is witnessing significant advancements with the integration of large language models (LLMs) and semantic-based ranking methods. Researchers are focusing on developing more accurate and efficient paper retrieval frameworks that can capture fine-grained scientific concepts and query intents. The use of neuro-symbolic methods, such as logical consistency and first-order logic, is also being explored to improve retrieval performance, particularly for complex queries like negative-constraint queries. Furthermore, hybrid retrieval pipelines that combine lexical precision, semantic generalization, and deep contextual re-ranking are being proposed to bridge the informal-to-formal language gap in scientific literature retrieval. Noteworthy papers in this area include:

• SemRank, which proposes an effective and efficient paper retrieval framework that combines LLM-guided query understanding with a concept-based semantic index.
• NS-IR, which introduces a neuro-symbolic information retrieval method that leverages first-order logic to optimize the embeddings of naive natural language by considering logical consistency between queries and documents.