Ryann829/SconeEval
收藏Hugging Face2025-12-17 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/Ryann829/SconeEval
下载链接
链接失效反馈官方服务：
资源简介：
SconeEval是一个新的基准测试，旨在评估模型在复杂视觉上下文中区分和生成指定主题的能力。它包含409个测试案例，涵盖字符、对象和场景组合以及主题区分，共19种案例类型和6个子任务。基准测试包括三个逐步挑战性的任务：组合、区分以及区分与组合。任务涉及跨类别和类别内的情况，要求模型在参考图像和指令的复杂上下文中进行跨模态推理。
SconeEval is a new benchmark designed to evaluate a model’s ability to distinguish and generate the referred subject in complex visual contexts. It contains 409 test cases across character, object, and scene combinations and subject distinction, with 19 case types and 6 subtasks. The benchmark includes three progressively challenging tasks: composition, distinction, and distinction & composition. Tasks involving distinction include cross-category and intra-category cases, requiring cross-modal reasoning from complex contexts involving reference images and instructions.
提供机构：
Ryann829
