Data and images for multimodal LLM conformity simulations
收藏官方服务：
资源简介：
images/ contains all generated image datasets for the three tasks (lines, color recognition, dots), including both general pools and perplexity-based difficulty/ambiguity sets, organized by task and model. data/ contains the corresponding experiment outputs (intermediate artifacts and summaries such as pickles/CSVs) produced by running the general notebooks over those image sets, organized by task and model.
提供机构：
Zenodo创建时间：
2025-12-22
