BioPhys-Bridge: A Benchmark for Interdisciplinary Scientific Reasoning in Physics-Grounded Biological Research

BioPhys-Bridge: A Benchmark for Interdisciplinary Scientific Reasoning in Physics-Grounded Biological Research

BioPhys-Bridge:一个用于物理驱动生物学研究中跨学科科学推理的基准测试

Language models face unique challenges in analyzing interdisciplinary scientific research literature. In biophysics research, faithful answers require grounding observed data in source evidence, interpreting it through a quantitative physics model, and linking it to a biological mechanism. 语言模型在分析跨学科科学研究文献时面临着独特的挑战。在生物物理学研究中,要获得准确可靠的回答,需要将观测数据建立在原始证据的基础上,通过定量物理模型进行解读,并将其与生物学机制联系起来。

To address this challenge, we introduce BioPhys-Bridge, a novel benchmark dataset for evidence-grounded scientific reasoning over biophysical literature. Each case contains evidence blocks, stable evidence IDs, quantitative values, units, equations, assumptions, mechanisms, and next decisions as grounding targets for question answering (QA) and retrieval-augmented generation (RAG). 为了应对这一挑战,我们推出了 BioPhys-Bridge,这是一个用于生物物理文献中基于证据的科学推理的新型基准数据集。每个案例都包含证据块、稳定的证据 ID、定量数值、单位、方程、假设、机制以及后续决策,作为问答 (QA) 和检索增强生成 (RAG) 的基础目标。

The initial release contains 500 cases, 1,517 agent-facing tasks, and covers six biological domains and nine physical model families, including three sparse families reserved for future expansion. We enforce strict quality gates for all cases in schema, evidence-integrity, quantitative-grounding, source-license, duplicate, unit-normalization, with domain expert review and annotation for 81 cases. 首个版本包含 500 个案例、1,517 个面向智能体的任务,涵盖了六个生物学领域和九个物理模型族,其中包括三个预留给未来扩展的稀疏模型族。我们对所有案例在模式、证据完整性、定量基础、来源许可、去重和单位标准化方面实施了严格的质量控制,并由领域专家对 81 个案例进行了审查和标注。

Preliminary evaluations show that DeepSeek-V4-Flash obtain the highest evidence-ID $F_1$ score (0.360), followed by Qwen3.7-Max (0.316) and GPT-4o-mini (0.294). BioPhys-Bridge is an interdisciplinary benchmark for evaluating attribution, faithfulness, hallucination reduction, and biological experiment design with complex, multi-step scientific reasoning. 初步评估显示,DeepSeek-V4-Flash 获得了最高的证据 ID $F_1$ 分数 (0.360),其次是 Qwen3.7-Max (0.316) 和 GPT-4o-mini (0.294)。BioPhys-Bridge 是一个跨学科基准测试,旨在评估复杂、多步科学推理中的归因、忠实度、幻觉减少以及生物实验设计能力。

Future works will increase the size and complexity of the dataset and perform comprehensive evaluations. Code and data are available in the GitHub repository and on Hugging Face. 未来的工作将增加数据集的规模和复杂性,并进行全面的评估。代码和数据现已在 GitHub 仓库和 Hugging Face 上发布。