SciToT：面向科学论文的多模态舌尖检索基准
原标题：SciToT: Multimodal Tip-of-the-Tongue Retrieval for Scientific Papers
AI 摘要
研究者提出 SciToT，一个面向科学论文的多模态「舌尖现象」（ToT）检索基准，包含 108,512 篇开放获取论文和 11,499 条合成查询。他们设计了一条可扩展的查询生成流水线，从文档中分别提取文本与视觉记忆线索，并用大语言模型生成类似人类不完整回忆的查询。该工作填补了科学文献 ToT 检索评测资源的空白，并以 MIT 许可开源数据集、文档表示、记忆线索、评测划分及生成与检索流水线。
正文节选
SciToT: Multimodal Tip-of-the-Tongue Retrieval for Scientific Papers Abstract Researchers frequently encounter difficulties re-finding scientific papers when they remember aspects of the content but cannot recall reliable identifiers, such as titles or author names. This type of re-finding task, known as tip-of-the-tongue (ToT) retrieval, is particularly challenging for scientific documents, where recollections may involve textual information and visual elements, including figures, diagrams, and