neulab/VisualPuzzles
收藏Hugging Face2026-03-27 更新2025-05-31 收录
下载链接:
https://hf-mirror.com/datasets/neulab/VisualPuzzles
下载链接
链接失效反馈官方服务:
资源简介:
VisualPuzzles是一个专门设计用来评估大型模型推理能力,同时故意减少对特定领域知识依赖的多模态基准。包含1168个不同的谜题,分为算法、类比、演绎、归纳和空间五种推理类别,难度分为容易、中等和困难。
VisualPuzzles is a multimodal benchmark specifically designed to evaluate the reasoning abilities of large models while deliberately minimizing reliance on domain-specific knowledge. It includes 1168 diverse puzzles across five reasoning categories: Algorithmic, Analogical, Deductive, Inductive, and Spatial, with difficulty levels of Easy, Medium, and Hard.
提供机构:
neulab



