WEAVE

🧵 WEAVE 📌 This is the official repository for the dataset presented in:Weave: A Benchmark for Evaluating Multimodal Editing Models 📊 Dataset Structure WEAVE consists of two main components: 🔹 WEAVE-100k (Training Set): Stored in data/ folder in parquet format 🔹 WEAVEBench (Test Set): Stored in test/ folder in both zip and json formats 🚀 WEAVE-100k WEAVE-100k is generated through four sophisticated pipelines and multiple validation stages… See the full description on the dataset page:

Source: Hugging Face Hub (WeiChow/WEAVE). Metadata imported from the dataset’s Hub tags.