BioTool: A Comprehensive Tool-Calling Dataset for Enhancing Biomedical Capabilities of Large Language Models
:::info Stub — Full Engineering Breakdown Coming This paper was featured on Hugging Face Daily Papers on 2026-05-07 with 2 upvotes. A full breakdown with production viability rating, implementation notes, and honest limitations is being written. Subscribe to AI Letters → :::
| Authors | Xin Gao et al. |
| Year | 2026 |
| HF Upvotes | 2 |
| arXiv | 2605.05758 |
| Download | |
| HF Page | View on Hugging Face |
Abstract
Despite the success of large language models (LLMs) on general-purpose tasks, their performance in highly specialized domains such as biomedicine remains unsatisfactory. A key limitation is the inability of LLMs to effectively leverage biomedical tools, which clinical experts and biomedical researchers rely on extensively in daily workflows. While recent general-domain tool-calling datasets have substantially improved the capabilities of LLM agents, existing efforts in the biomedical domain largely rely on in-context learning and restrict models to a small set of tools. To address this gap, we introduce BioTool, a comprehensive biomedical tool-calling dataset designed for fine-tuning LLMs. BioTool comprises 34 frequently used tools collected from the NCBI, Ensembl, and UniProt databases, along with 7,040 high-quality, human-verified query-API call pairs spanning variation, genomics, proteomics, evolution, and general biology. Fine-tuning a 4-billion-parameter LLM on BioTool yields substantial improvements in biomedical tool-calling performance, outperforming cutting-edge commercial LLMs such as GPT-5.1. Furthermore, human expert evaluations demonstrate that integrating a BioTool-fine-tuned tool caller significantly improves downstream answer quality compared to the same LLM without tool usage, highlighting the effectiveness of BioTool in enhancing the biomedical capabilities of LLMs. The full dataset and evaluation code are available at https://github.com/gxx27/BioTool
Engineering Breakdown
Plain English
This paper presents BioTool, a dataset specifically designed to teach LLMs how to call biomedical tools and APIs from major databases like NCBI, Ensembl, and UniProt. The dataset contains 34 frequently-used biomedical tools and is meant for fine-tuning LLMs to better handle specialized biomedical tasks, addressing a gap where existing models struggle in this domain compared to general-purpose applications.
Key Engineering Insight
The core insight is that biomedical LLMs fail not because they lack domain knowledge, but because they can't reliably invoke the specific tools that domain experts actually use in their workflows. Fine-tuning on real tool-calling examples is more effective than relying on in-context learning or restricting models to small tool sets.
Why It Matters for Engineers
If you're building AI systems for clinical or biomedical research workflows, your LLM needs to actually call databases and tools—not just talk about them. This dataset gives you a training resource to make models that can programmatically query NCBI, retrieve gene sequences, or look up protein annotations, turning them from conversational systems into functional agents that integrate into real research pipelines.
Research Context
General-domain tool-calling datasets have successfully improved LLM agents, but biomedical applications were stuck using limited in-context learning approaches with small tool sets. BioTool moves biomedical LLMs to the same level by providing comprehensive fine-tuning data, enabling models that can operate as real assistants within specialized scientific workflows rather than just providing information retrieval.
:::tip Subscribe Get weekly breakdowns of papers like this in AI Letters - the newsletter for engineers building production AI systems. :::