LLMs grounded in curated biomedical databases for gene and literature reasoning

Published papers

• MedHopper: an agentic RAG-LLM system for multi-hop biomedical QA.

  Database : the journal of biological databases and curation2026-01-15Rustam Ruslanovich Taktashov et al.

  MedHopper: agentic RAG-LLM with bounded state machine and reranking for multi-hop biomedical QA, scoring 0.55 EM on MedHopQA.

• I-NOWJ: adaptive context selection and knowledge-guided learning for substance use named entity recognition.

  Database : the journal of biological databases and curation2026-01-15Quoc-Viet-Anh Tran et al.

  I-NOWJ uses knowledge-guided augmentation, self-training and adaptive context for Spanish clinical substance-use NER (ToxHabits).

• MedHopQA: a disease-centred multi-hop reasoning benchmark and evaluation framework for LLM-based biomedical question answering.

  Database : the journal of biological databases and curation2026-01-15Rezarta Islamaj et al.

  Evaluating large language models (LLMs) in the biomedical domain requires benchmarks that can distinguish genuine reasoning from pattern matching and that remai

• Federated knowledge retrieval elevates large language model performance on biomedical benchmarks.

  GigaScience2026-01-21Janet Joy et al.

  BTE-RAG augments LLMs with retrieval from a federated API of over sixty biomedical knowledge sources, improving accuracy on gene, metabolite, and…

Preprints

• From Single to Multi-Agent Reasoning: Advancing GeneGPT for Genomics QA

  arXiv2026-01-15Kimia Abedini et al.

  Proposes GenomAgent, a multi-agent LLM framework for genomics QA that coordinates specialized agents over databases; reports 12% average gain over GeneGPT on…

• SciHorizon-GENE: Benchmarking LLM for Life Sciences Inference from Gene Knowledge to Functional Understanding

  arXiv2026-01-19Xiaohan Huang et al.

  SciHorizon-GENE is a large gene-centric benchmark of over 540K questions from curated databases (190K+ human genes) evaluating LLM gene-to-function reasoning.

• RAGulate: Retrieval-Augmented Generation for Post-hoc Literature-Grounded Regulatory Assessment

  bioRxiv2026-01-21Zandigohar, M. et al.

  RAGulate: a retrieval-augmented LLM framework using CollecTRI and PubMed evidence to assess TF-target regulatory predictions in context, with reduced…

Related topics

• Llm Agents For Drug Discovery
• Agentic Llms For Omics Analysis
• Agentic Llm Systems For Biology

Topic history