trillionlabs/TheBioCollection-Eval
收藏资源简介:
TheBioCollection-Eval是一个生物学评估套件,用于评估生物领域大型语言模型(BioLMs)在多个生物学子领域的性能。它覆盖小分子、蛋白质、基因组序列、细胞/通路和跨领域推理等方向,通过整合现有基准(如Mol-Instructions、MolLangBench、BioReason-Pro、PerturBench Replogle K562)和新构建的指令数据集构建而成。套件包含总计1650个查询,涉及描述引导的分子设计、蛋白质功能预测、基因组序列定位、细胞类型分类、跨领域推理等多种子任务,数据以压缩JSONL格式存储,专为研究评估设计,不适用于临床或治疗决策。
TheBioCollection-Eval is a biological evaluation suite for assessing large language models (BioLMs) for biology across small molecules, proteins, genomic sequences, cells/pathways, and cross-domain reasoning. It is constructed by drawing subtasks from many scattered existing benchmarks (Mol-Instructions, MolLangBench, BioReason-Pro, PerturBench Replogle K562) and combining them with source-derived newly-constructed instruction datasets. The suite includes a total of 1,650 queries covering various subtasks such as description-guided molecule design, protein function prediction, genomic sequence localization, cell-type classification, and cross-domain reasoning. Data is stored in compressed JSONL format and is intended for research evaluation only, not for clinical or therapeutic decision-making.




