harshal3099/apex-food-rd-chatml
收藏资源简介:
Apex食品配方研发ChatML数据集是一个合成的监督微调数据集,专注于印度清洁标签功能性食品的配方研发。数据集包含5,500个ChatML示例,分为训练集4,950个、验证集275个和测试集275个。格式为包含system、user和assistant的messages列。数据集涵盖了多种能力分布,如成分功能、监管许可、保存系统设计、清洁标签替代、工艺工程、FSSAI和欧洲标准健康/营养声明、ICMR-NIN饮食差距分析、有机认证、发酵科学、保质期预测以及质地和感官设计。数据集基于印度和欧洲的监管/科学知识库生成,并引用了多个法规和标准。需要注意的是,这是一个合成的MVP训练数据集,主要用于教学响应结构、配方推理和引用行为,在用于商业合规决策前,应验证答案的准确性。
The Apex Food Formulation R&D ChatML Dataset is a synthetic supervised fine-tuning dataset focused on Indian clean-label functional foods for Apex Nutrition. It contains 5,500 ChatML examples, divided into 4,950 for training, 275 for validation, and 275 for testing. The format includes messages column with system, user, and assistant. The dataset covers various capability distributions such as Ingredient Functionality, Regulatory Permissibility, Preservation System Design, Clean-Label Substitution, Process Engineering, FSSAI & European Standards Health/Nutrition Claims, ICMR-NIN Dietary Gap Analysis, Organic Certification, Fermentation Science, Shelf-Life Prediction, and Texture & Sensory Design. The dataset is generated from a curated regulatory/scientific knowledge base and cites multiple regulations and standards. Important limitation: This is a synthetic MVP training dataset designed to teach response structure, formulation reasoning, and citation behavior. Before using the model for commercial compliance decisions, validate answers against current source PDFs and legal/regulatory counsel.




