ButterChicken98/soyabean_aerial_plus_healthy_rag_short_ablation

该数据集包含1400个训练样本，总大小约为109.8 MB。每个样本包括图像、标题、标签以及多种提示相关字段，如随机提示、检索提示和RAG（检索增强生成）提示，用于支持多模态任务。数据特征涵盖图像、字符串、浮点数和布尔类型，具体包括image（图像）、caption（标题）、captions（标题列表）、label（标签）、prompt_random（随机提示）、prompt_random_id（随机提示ID）、prompt_retrieval_k1（检索提示）、prompt_retrieval_k1_id（检索提示ID）、prompt_retrieval_k1_score（检索提示分数）、prompt_retrieval_k1_aspect（检索提示方面）、prompt_rag_k2_slm（RAG提示）、prompt_rag_k2_slm_ids（RAG提示ID列表）、prompt_rag_k2_slm_scores（RAG提示分数列表）、prompt_rag_k2_slm_aspect（RAG提示方面）、prompt_rag_k2_slm_fallback（RAG回退标志）、rag_short_class_key（RAG短类键）和rag_short_source_corpus（RAG短源语料）。数据集仅提供训练分区，适用于图像描述、分类和生成任务。This dataset contains 1400 training samples with a total size of approximately 109.8 MB. Each sample includes images, captions, labels, and various prompt-related fields such as random prompts, retrieval prompts, and RAG (Retrieval-Augmented Generation) prompts to support multimodal tasks. The features cover image, string, float64, and boolean types, specifically including image, caption, captions, label, prompt_random, prompt_random_id, prompt_retrieval_k1, prompt_retrieval_k1_id, prompt_retrieval_k1_score, prompt_retrieval_k1_aspect, prompt_rag_k2_slm, prompt_rag_k2_slm_ids, prompt_rag_k2_slm_scores, prompt_rag_k2_slm_aspect, prompt_rag_k2_slm_fallback, rag_short_class_key, and rag_short_source_corpus. The dataset only provides a training split and is suitable for image captioning, classification, and generation tasks.